[Research Project] “The Opioid Crisis is currently one of the most significant crises occurring in America today … [read more] According to the NIH, over 9 million Americans misuse opioids annually, over 5.5 million Americans live with opioid use disorder, and over 100,000 die from drug overdoses, with opioids being involved in 75% of those deaths. To help address this crisis, Machine Learning (ML) models will be used to predict potential locations that may be a future opioid outbreak point by analyzing demographic trends and geographic potential locations that may be a future opioid outbreak point by analyzing demographic trends and geographic patterns in opioid use in United States. Furthermore, genetic factors that are associated with opioid addiction, such as specific dopamine receptor alleles, will be used for diagnosing people with risk for potential drug opioid addiction. For this project, two machine learning models will be used. Logistic Regression will be used to predict the likelihood of a location becoming a future opioid outbreak location and the risk for a person developing an opioid addiction. Along with Logistic Regression, Neural Networks will be used to score the likelihood of an opioid outbreak occurring at hte inputted location as wll as the likelihood of an individual developing opioid addiction. In order to measure the performance of the machine learning models, F1 Score, Precision Score, and Accuracy Score will be used to prevent the consequences of both false positives and false negatives and provide a general measure of the mosel’s performance.” [/read]

Presentation Click Here

[Research Project] “Substance and alcohol abuse are one of the biggest issues in modern society that lead to fatalities … [read more] Current treatment for these habits consists of rehabilitation centers and care from health professionals. Sometimes, patients are successful with leaving their old habits and becoming a normal individual, but what happens when outside factors lead them to repeat their past habits? That is called relapse, and unfortunately, relapse rates are especially high with about 70% and 85% of individuals with a past record of alcohol or drug abuse relapsing within a year of treatment in the US. My proposal comes from being able to minimize this relapse rate through the use of machine learning modeling. Necessary data metrics include age group, time since relapse, gender, and current employment status that can all be used to for a single-time model such as random forest classifier. Other data metrics that require human feedback and input, such as mental health history and type of relapse they are prone to, utilize a personalized app that takes in the responses and performs natural language processing, sentimental analysis, and learning algorithms like support vector machine. These proposals would be effective if combined together, or separate, since gaining statistical and/or word inputs can maximize the assessment of a high chance that someone may relapse and be sent to a call landline immediately to address the issue.” [/read]

Presentation Click Here

[Research Project] “For my research project, my idea of how we could use Al to resolve or mitigate the issue of Drug Abuse … [read more] is meditation therapy. Most of the time when patients take medicine, it is highly likely that they become overly dependent and eventually addicted to the drug. The main issue of drug addiction and abuse is the fact that many patients can not stop the use of drugs completely as many people still need it for their medical issues. So even if you can’t keep patients from stopping the use of drugs completely, we can use meditation therapy, so that patients use minimum amounts and depend on it less. So where does AI fit into this? AI can identify if a Patient is extremely drug dependent by analyzing their face, voice, or any behaviors. By using AI instead of real-life people therapists, more people can be treated at the same time. This means, no more waiting for your turn at therapist’s offices. You can simply grab an electronic device and help is there in your hands. By Al targeting a larger population at once, not only does it cover the issue of time, but it is also more cost-effective and solves the issue of a shortage of therapists. For general background information, I would need information on the early symptoms of drug abusers or addicts, how accurate Al has been in the past in analyzing facial recognitions or behaviors, and statistics on meditation. For personal information from the patients. I would need patient profiles: medical history, family background, and any information of the patient that could be a cause factor of their addiction. Next, video analysis, and Al’s understanding of emotions and sentimental issues.
Lastly, two possible machine learning models that could be used to accomplish my project idea are emotion recognition. This uses face recognition and voice analysis to analyze facial expressions. tone of voice. and sentiment. Additionally, reinforcement learning, which gives Al the ability to have adaptive responses allows Al to ask questions in their sessions, and depending on the patient’s responses, it would lead to different questions, activities, and comments. This idea is similar to decision trees. Overall, this makes Al therapy to be more personalized care for each person.” [/read]

Presentation Click Here

[Research Project] “My project was centered around how social media platforms could be used to prevent drug overdose deaths but more … [read more] importantly, prevent new users. It would function by training an AI model off how online behavior changed as people took up drugs and use that to detect new possible intakers. Based on research by institutions, Ifound that the majority of teenagers who are likely to start drugs and alcohol have a social media profile with many having an active account. This means that the vast majority of new consumers would show a change in their posts or interests as they transitioned. In order to analyze these changes, a model that could interpret language and images would be necessary. Therefore, I settled with a deep learning strategy called BERT. It is able to understand the context of words and sentences very well which made it an excellent choice. For images, I went with CNN. This was because it was something that we had already discussed and learnt during the lectures and it was efficient and accurate. The data required to train these models are widely accessible but would take a considerable amount of time to collect and organize. If this project was to be actually put into action, another AI model could potentially be used to automate the sorting and remove manual labor. Lastly, for the performance metric, I picked precision and recall score because they ran on the same basis. However, the recall score would likely be more important in this case as a false negative would be quite detrimental as opposed to a false positive which would only really result in embarrassment.   ” [/read]

Presentation Click Here

[Research Project] “The relationship between genetics and substance abuse still remains very poorly understood … [read more] In 2022, 70.3 million people (aged 12 or older) used prohibited drugs with the most common drug being marijuana used by 61.9 million people (aged 12 or older).(Substance Abuse and Mental Health Services Administration) With very few studies exploring the effect genetics may have on substance abuse, using an AI model could examine and help us understand the associations between genetic factors and substance abuse. However a study showed conclusions for some genes such as DRD2, that may have a significant impact on substance abuse which will be focused on when training the AI model.(Hatoum et al.) The AI will be trained with a dataset containing genetic information as well as their diagnosis for substance abuse disorder. The AI will utilize two different models: logistic regression and k-nearest neighbors (kNN). Using logistic regression, the model will find the weights of each gene, predicting an individual’s risk of substance abuse disorder. The kNN model will find similarities in genes between the training data set with an individual’s genetic information, also outputting the risk of substance abuse disorder. This AI proposal will use an individual’s genetic information to quickly provide their risk for addiction and allow earlier access to help. Ultimately, this highlights the potential of AI to help us understand the intricate contributions genes may have on substance abuse and obtain quick insights to help subdue this health crisis. Citations:Substance Abuse and Mental Health Services Administration. (2023). Key substance use and mental health indicators in the United States: Results from the 2022 National Survey on Drug Use and Health (HHS Publication No. PEP23-07-01-006, NSDUH Series H-58). Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration.https://www.samhsa.gov/data/report/2022-nsduh-annual-national-report Hatoum, Alexander S., et al. “Multivariate Genome-Wide Association Meta-Analysis of over 1 Million Subjects Identifies Loci Underlying Multiple Substance Use Disorders.” Nature News, Nature Publishing Group, 22 Mar. 2023, www.nature.com/articles/s44220-023-00034-y#Sec19.” [/read]

Presentation Click Here

[Research Project] “My research project is titled “Identifying the type of drug in a person’s system for faster … [read more] treatment.” For drugs, there are several different categories that exist. For example, there are stimulants, depressants, opiods, psychedelics, empathogens, etc. Each category comes with its own unique symptoms and side effects, meaning they require different treatment plants and overdose treatment drugs. Therefore, for health professionals, including doctors, nurses, or therapists who work with substance overdose or abuse patients, it is incredibly important to quickly identify the category of drug that their patient is using to formulate the best course of treatment and avoid any unwanted side effects. To do this, I decided to apply AI to two data sets. The first one is one that needs to be created. It will include the blood test results of different drug patients, including the cell, protein,  and oxygen concentration. This way, a K-Nearest Neighbors model can be run on this data and the test point to rapidly categorize the drug type from a quick blood test. Another set, called “Current Final Multiple Cause of Death Data,” can be provided by the CDC. It contains the demographic information of different drug patients, allowing us to use a neural network to compare the demographics (age, gender, state of residence, ethnicity, etc.) of the patient to known train data in order to categorize the most likely drug used. Through these methods, we will be able to efficiently diagnose and treat substance abuse patients.” [/read]

Presentation Click Here

[Research Project] “My research for the final project presentation was focused on where drug-related deaths … [read more] occurred the most Drug abuse is a huge problem here in the states, but the government has limited resources to tackle the situation. Thus, I decided that a machine learning model which could identify areas with high ‘risks’ of drug abuse would be useful for classifying which areas these resources should be concentrated. To find detailed data sets that included information about the specific location of a drug abuse incident, I researched data online and found a detailed dataset from the Cook County Medical Examiner Case Archive. This dataset included specific coordinates and details of every death which has occurred during a specific time period in Cook County, Illinois. I figured that such data would be ideal for a neural network model. With just the right weights and bias, I figured that I could use filtered, preprocessed data from the dataset to train a neural network model that can predict the risks of drug abuse in specific neighborhoods. I also proposed the use of a CNN model trained with Google Maps Satellite Image data for visualizing areas where the risks of drug abuse are high. Finally, I suggested the use of precision scores to evaluate my model. ” [/read]

Presentation Click Here

[Research Project] “Substance drug abuse is currently one of the biggest issues in America. According to the NIDA … [read more] common drugs used include illicit drugs, synthetic opioids, and pain relievers. Misusing these drugs can lead to having poor relationships with loved
ones, addiction, and in the worst cases, death. Unfortunately, drug abuse deaths have only increased over the years. According to the CDC, over one million people have died from a drug overdose since 1999, with there being 106,699 drug overdose deaths in 2021 alone. To help address this issue, my research project explained how two different machine learning (ML) models could be used to help people struggling with drug addiction before it was too late. My research project was titled Drug Abuse and Machine Learning: Combating the Drug Epidemic . In it, I described how two machine learning models, logistic regression and decision trees, could be used to prevent the misuse of drugs. Logistic regression is a machine learning model with binary classification. Using data with factors such as race, income status, age, and sex of the user, the model can predict the probability rate of the user misusing substance drugs. Additionally, we can use preprocessing to remove missing values and encode categorical information in the data sets. A decision tree is a machine learning model Presenter that can be used to take inputs from a user and output decisions that are not just ‘yes’ or ‘no’ Decision trees can use data to find different patterns of drug abuse and using these patterns, take inputs given by the user and output decisions based on the user’s responses. Lastly, the best performance metric for both of the machine learning models mentioned above would be recall, which can capture all positive instances. In other words, while it may cause false alarms, it would be best at detecting a user falling victim to drug abuse.” [/read]

Presentation Click Here

[Research Project] “My project was on stress and anxiety on drug abuse. My project is to use CNN facial scanners … [read more] to detect stress and anxiety and use those outputted values as input for an SVM model to predict likeliness of substance abuse. Data for facial scan would be from images of individuals and their expression under categories of stress and anxiety. This is public data for stress and anxiety and substance abuse likeliness. My preprocessing data would be to take out outliers and inefficient data. To elaborate, my
preprocessed data will include thousands of images assigned with two values, stress and anxiety, through a 0-1 scale. Example, imagine a woman very happy. Her values will be [0,0], has no stress or anxiety. Now, imagine a man very stressed. His values would be [0.9,1], very stressed and very anxious. I will train my SVM using data from resources and split my preprocessed data into 80% training and 20% testing when training the model. Using ML models, I will use CNN and SVM. For CNN, the input will be facial images and the output would be the two numbers for stress and anxiety (the assigned values). For SVM, the input would be the assigned values for stress and anxiety from CNN output. The output for SVM would
be the likeliness of substance abuse for the individual. The performance metric of this project is accuracy. The idea is by predicting likeness to substance abuse based on expression, one gets treated to prevent future incidences.” [/read]

Presentation Click Here

“The class was fun at the beginning but towards the end, I thought it was stressful … [read more] Having to speak in a zoom call full of kids I have never met was hard for me. The teachers were amazing at teaching, but I learn a different way than just listening and looking. They gave us a coding sheet to work on which helped a lot for me to focus. The virtual setup made it hard, but having something hands-on like the coding sheet was a game-changer. It made the class less boring and more practical. The teachers did their best, but the Zoom thing just didn’t work great for me. I liked that the coding sheet gave me something to do instead of just staring at a screen. It made the class less stressful and more about doing stuff, which is how I learn best.” [/read]

[Research Project] “My project aimed to tackle the pressing issue of drug abuse, a problem affecting countless lives … [read more] particularly among individuals as young as 17. I proposed utilizing computer vision as a preventive measure. The concept involved a machine assessing a person’s well-being by asking about their day. If signs of distress or abnormalities, such as a frown or weakened eyes, were detected, or if the person deviated from their usual behavior, the machine would intervene. It would alert authorities and halt any potentially harmful actions. The motivation behind my project was the alarming rate of overdoses among teenagers in the United States. Knowing that peers just a couple of years older than me were succumbing to depression or succumbing to drug pressure deeply saddened me. I believed my project had the potential to save lives, but I acknowledge that it required more dedication to coding for full implementation. Preventing these deaths became a personal mission. The idea of peers losing their lives due to societal pressures and mental health struggles fueled my determination to create a solution. Although my project remained at an ideational stage, the gravity of the issue and the potential impact of my proposal added a sense of urgency to my commitment to addressing drug abuse.” [/read]

[Research Project] “Drug abuse is an immense problem all around the world, especially in the US. Identification is the first … [read more] step toward preventing and reducing drug abuse. We can use data, such as location, median income, gender, etc.to predict and prevent possible future deaths. Although the predictions will not be perfectly accurate, they will help us indicate areas where assistance is vital. For my research project, Ifirst researched data sets about drug abuse deaths. With these data sets, Ican use machine learning to train my model to predict future possible deaths. The data set Iwill mainly be using Is on drug abuse deaths in Connecticut and connect this to the median household income and the population. The next step is to predict, which is mostly done by the Al model. After training the machine learning model, it can easily predict cities with major drug abuse problems. With the trained model, I will be able to input factors of different cities to find their probability of drug use and classify them into different degrees of drug abuse risk.. To increase the accuracy of my predictions, I will input additional variables and data from other cities. The final step is to prevent, arguably the most important step. One major way I plan to do this is to spread education via flyers and volunteers throughout high-risk cities and neighborhoods. In conclusion, Ihave learned so much about machine learning and drug abuse prediction and prevention around the world.” [/read]

[Research Project] “My presentation, Information Handling Regarding Drugs and ML, aims in combating … [read more] the substance abuse and alcohol problem was through the usage of a chatbot. Furthermore, the problem is that not everyone is aware of the consequences of certain drugs and safe dosage; how can we control information and use it to educate those who buy drugs? Briefly, I created a linear regression model and a Random Forest Model to create the general
function of the chatbot. When a user goes on a platform to buy certain drugs or products containing certain chemical compounds, the chatbot would appear and note them about the statistics regarding the drugs and allows for interaction between the user. Within my Random Forest Model, if a user inputs, after purchasing their desired drug, that they know the risks of the drug, then the bot will then assume that the user knows and will suggest the drug’s best uses, and how to use it safely. If the user inputs that they don’t know the risks, the chatbot will input that they’re at a risk of misusing and that overdose, possible death, and their lives are at stake. Thus, it will give statistical data regarding the mortality rate of the drug after overdose, the frequency, and how to use it safely as well. The point being to spread as much awareness using machine learning models which are used within the chatbot.” [/read]

[Research Project] “This project is all about using smart technology to tackle the issue of drug abuse. We’re exploring … [read more] how artificial intelligence (AI) can step in to help out. By bringing in psychological data, we aim to stay ahead in predicting and spotting the factors linked to drug abuse. It’s like having a heads-up to prevent the widespread impact of this problem. We’ve got a mix of data sets that we can use here. From early identification using surveys and interviews to figure out people dealing with stress or mental health issues, to tailoring educational programs that tackle specific stuff like low self-esteem or peer pressure. The idea is to create personalized stragegiers that fit each person’s unique situation. We’re also looking at risk prevention models, which mash up psychological data with other factors. This helps us build predictive models to figure out who’s at risk of drug abuse. It’s like aiming our efforts where they’ll make the most impact. Here are the details of handling the data. We’re doing things like spotting outliers standardizing numbers, and keeping things balanced. Privacy is also a big factor as well. Then, we’ve got two types of smart programs-logistic regression and neural networks. Logistic regression is like a yes-or-no algorithm, while neural networks can handle more complex relationships in different kinds of data. When it comes to checking how well these models work, we’ve got metrics like precision, recall, and F1 score. Basically, we want to make sure we’re catching the right cases without causing mistakes. It’s about finding the most accurate answer and making sure we don’t miss anything important. In this case, making sure we don’t miss out on spotting potential drug abuses cases is key.” [/read]

[Research Project] “My research project explored the correlat ion  between poverty rates and drug abuse in America … [read more] emphasizing the application of artificial  intelligence (Al) to analyze statistical data effectively. Two datasets from credible sources were  used. The first dataset provided information on the total deaths and percentage of drug-related deaths for each state. The features allowed me to analyze the trend correlation with each state’s population and income level. The second dataset contained information about each state’s total number and percentage of people in poverty. By employing two adva nced machine learning algorithms, logistic regression and k-nearest neighbors, the study aimed to discern patterns and trends in every state. Logistic regression predicted a probability of an event occurring. With the raw data of poverty rates and demographics, the model classified individuals’ probability of drug abuse. The KNN model classified data points based on the majority of the k-nearest neighbors in the featured area. The accuracy, precision and Fl scores were used to evaluate the model performance with the visualization of a confusion matrix. Accuracy provided a view of how well the model classified areas of high and low risk. Precision assessed a model’s accuracy in locating areas, and F l provided a balanced assessme nt by considering both false positive and negatives. My resea rch underscored the potential of Al in creating a proximity map with real-time monitoring to show spatial clustering of areas with low poverty rates.” [/read]

[Research Project] “AI can play a crucial role in addressing drug abuse by predicting and preventing relapse … [read more] for certain individuals experiencing drug addiction. One of the ways we can utilize AI is through wearable devices using pattern and image recognition to stream real-time data to the two possible machine learning recognition models. I believe to accomplish this is through two models known as, Random Forest (RF) and Support Vector Machines (SVM), two powerful machine learning models used to process this data. RF is used for its high accuracy and visibility. Models such as RF and SVM can be used for texts, classification face detection, and more. The data I would personally use is the historical data on drug abuse cases, treatment outcomes, and the demographic information of the state in which I am doing my case. I would also utilize national surveys on drug use and health. Artificial Intelligence being utilized for substance misuse treatment is a significant breakthrough, especially in terms of anticipating and averting individual relapses. One creative way to use AI is through transmitting real-time data continuously through wearable technology, pattern detection, and even image recognition. The RF and SVM will then process this data. Furthermore, adding state-specific demographic data improves the geographical analysis and makes it easier to implement focused interventions and preventative measures. This strategy provides tailored, data-driven insights and interventions that improve treatment program efficiency, with the ultimate goal of preventing drug abuse at the individual and societal levels.” [/read]

[Research Project] “As mentioned above, our goal in this program was to develop a means of preventing drug … [read more] overdoses through the adaptive and innovative algorithms of Al. My research led me to a solution that utilized both medical data and patient history inputted into two different algorithms, namely K Nearest Neighbors and Logistic Regression. K Nearest Neighbor is a machine learning algorithm in which the system detects patterns and trends within the data to make predictions regarding the concern. Logistic regression converts data sets into probabilities which are then used in a risk factor score to determine whether someone is at risk or not. Both algorithms utilize data and an algorithm to predict future events such as the chances of a drug overdose or who is most likely to overdose. By using, medical data, these algorithms would be able to determine the overdose risks for millions over a widespread range of demographics and geographical locations. In this study, I also utilized the RA2 metric and the precision metric to gauge the accuracy of these algorithms as well as their margins of error. This allowed for a smooth study that was able to prevent drug overdoses. I concluded that machine learning was indeed an effective means of preventing drug overdoses by utilizing medical data and records.” [/read]

[Research Project] “As stated, the goal of the program was to create an adaptive algorithm using machine … [read more] learning to prevent drug overdoses My research allowed me to come with an idea in which I would utilize the patient’s demographic, medical history, and more to create a model that would predict drug usage and allow early intervention. I used two methods called linear regression and decision trees. Linear regression is a machine learning algorithm that takes multiple points on a graph and predict a line of best fit. Decision trees starts with a single block which would ask a question that leads into multiple different blocks based on how the question is answered. Both of these models utilized data and were algorithms that would be able to determine the overdose risks for millions over a widespread range of demographics and geographical locations. In this study, I also utilized the R^2 metric and the precision metric to gauge the accuracy of these algorithms as well as their margins of error. This allowed for a very efficient and accurate prediction on overdoses. Machine learning was a very effective way of preventing drug overdoses by using a wide range of data on a patient.” [/read]

[Research Project] “My idea was very simple but very impactful. My initial idea was to target one singular state in … [read more] the United States and take specific towns/areas and research the amount of people doing alcohol and substance abuse. After looking at which areas I wanted to see what specific group of people would use substance abuse and I would calculate the average group of people in each town/area to see which group to look after and help. When I mean a specific group I am looking into all factors such as age, race, and gender. So for my project I looked at multiple sources to look for data about the area I was researching and while doing so I made a data list to calculate all my data. I looked at everything to have the most accurate data. I looked at the population of every town/area, deaths in the town/area, how much of each gender lives there, how much of each race lives there, and the deaths in each gender/race. Using this data list I could figure out a pattern from looking at each data and figure what specific group of people should usually need help in that area.” [/read]

[Research Project] “In addressing the alarming issue of substance abuse among teenagers, artificial intelligence … [read more] can be used as a helpful tool to assist vulnerable populations. The goal of this project was to apply AI in a way that we can identify and prevent substance use among teenagers. This approach blends technology with sensitivity, providing hope for communities and families. In my project, I used two different machine learning models to graph how AI can identify drug abuse among teenage populations. A machine learning model is a program that can identify patterns or make decisions from a previously unseen dataset. While working on my project as part of this course, I utilized a dataset that shows the high and low risks of different types of drugs and substances. Using the percentages of participates and the related risks, I created a data table to survey the usage patterns of different types of drugs and substances. By comparing past trends in drug usage with present trends, the goal is to be able to predict substance abuse patterns among youth populations. This project demonstrated the feasibility with which artificial intelligence can be applied to identify and manage health-related issues in our community on a broad scale.” [/read]

“The GLR-AI class, focused on the impact of artficial intelligence on substance abuse and alcohol … [read more] proved to be a multi-faceted learning experience for me. It not only imparted knowledge about various machine learning models but also shed light on the escalating issue of drug-related deaths in our society. Simultaneously, the class equipped me with the skills to propose and undertake projects. Prior to enrolling in this class, my understanding of AI and coding was limited. I possessed only basic knowledge of Python and could not have envisioned the potential of coding in addressing substance abuse. I used to believe that coding was solely for computing numerical values and lacked applicability to solving societal problems. However, this class challenged my perception and demonstrated how coding could play a crucial role. Exploring the diverse techniques employed by computers, such as image recognition and patern detection, filled me with excitement about the capabilities of machines. The class illustrated that, with the right techniques and purpose, machines have the potential to achieve anything. It dismantled my previous misconceptions about the limited scope of coding and opened my eyes to the transformative impact it can have on addressing societal challenges.” [/read]

[Research Project] “People of various ages in every country struggle with substance and/or alcohol abuse … [read more] In 2021, the United States reported 106,699 drug-involved overdose deaths. Drug overdose has emerged as a significant societal problem, and its impact on thousands of lives will persist without effective solutions. One such solution, working towards preventing and/or limiting substance abuse, involves leveraging AI technology. The utilization of machine learning models, such as convolutional neural networks, allows for the implementation of a required face scan capable of detecting individuals suffering from substance abuse worldwide. By analyzing the input of a person’s scanned face, a machine learning model can accurately determine whether the individual is experiencing drug abuse. This process relies on the recognition of paterns previously studied, enabling machines to differentiate between drug abusers and non-drug abusers by comparing pixels of our model image to those of the inputed image. This technique empowers computers to recognize differences in eye color, skin tone, eye size, and various other identifying features of a person struggling with substance abuse. Administering this face scan globally enables the identification of individuals in need of assistance with their addiction. Subsequently, mandatory therapy sessions or support groups can be provided to substance abusers, contributing to a reduction in drug-related  deaths.” [/read]

[Research Project] “My research project for this UNSLA program was regarding ‘Al to resolve drug overdosing.’ … [read more] I really enjoyed myself while making this project. First, I have the table of contents of the presentation and the driving question. Next, I did some research about the statistics of how many people overdose on which drug. Unfortunately, I was unable to find the exact data I was looking for; a clear and concise data set of overdoses organized by the specific drug in a graph was needed for my solution with Al. Then, I moved on to the actual models that I would use to solve my problem. My first model was a boxplot. I had the idea to use this model to visualize and organize data numbers by finding the median number for overdoses, which would help identify which drugs should be top priority when it came to getting it more restricted access. After that, I moved on to my second model to solve the problem. My second model I used was decision trees. The decision trees would be used to ask yes or no questions to determine which highly used overdose drugs should be prescription only drugs by doctors and be more highly restricted.” [/read]

[Research Project] “I proposed an innovative approach to combat drug abuse using artificial intelligence (AI) in … [read more] rehabilitation centers particularly for identifying drug overdoses quickly and accurately. The strategy employs a machine learning model trained on the “Accidental death associated with drug overdose in Connecticut from 2012 to 2018″ dataset from Kaggle, which contains 5,099 reports detailing race, age, cause of death, and location. These qualitative data points have been numerically transformed for AI processing. The model’s first phase utilizes Convolutional Neural Networks (CNN) to analyze camera footage from rehab centers, enabling the AI to identify patients’ race, age, and gender. This identification is crucial for the next step, which involves using Logistic Regression to determine the type of drug abuse. This method estimates the probability of various drug overdoses, facilitating a tailored and rapid response in administering the correct medication. For instance, the model might predict that a 25-year-old white female is most likely to have overdosed on heroin. The effectiveness of this AI model is measured using an accuracy score, a suitable metric for Logistic Regression’s probabilistic predictions. My approach aims to revolutionize the way rehabs and hospitals respond to drug overdoses, prioritizing quick and accurate treatment without relying on time-intensive lab tests. This model, specific to Connecticut’s data,  represents a targeted solution to drug abuse, potentially adaptable to other regions with unique datasets.” [/read]

“Delving into the GLR-AI class focused on the impact of Artificial Intelligence (Al) on substance … [read more] abuse and alcohol helped me understand how Al and machine learning helps to solve real-world problems. The course emphasizes practical application of algorithms and data to help prevent, intervene, and treat the issues of substance abuse. The ability of machine learning models to predict patterns of substance abuse for early identification and personalized interventions is a noteworthy aspect explored in the class. The inclusion of teachers is very nice, as they give accessible and stigma-free support to students trying to learn about machine learning. The GLR-AI class extensively covers the use of machine learning in substance abuse research and treatment methodologies.” [/read]

[Research Project] “My project was based on the prediction of drug abuse. I would use machine learning models … [read more] and Al tools taught to us during the class to predict who would be most susceptible to using drugs. Not only would I try to pinpoint who exactly would be most at risk for drug use, but also which communities and areas are at risk. I tried to use a K nearest neighbors machine model for this, as the K nearest neighbors model was the best for big data sets that still needed to be refined.” [/read]

[Research Project] “My research for the UNSLA was for the Drug abuse and how community can really help the people … [read more] that suffered from substance and drugs abuse. Community could really make Shelters for them and interventions the people that are victims of drugs abuse could get a second chance in life and get their life back again. While people with alcohol problems to make programs for them to refrain them both drug abuse and alcohol users to get withdraw and to get them in a straight path. I used some things from stigma showing all the possible ways artificial intelligence and Al could think of ideas and help refrain people from overdose and addictiveness. Al could really help the world in the future it could be good or bad but it will help people with better medicines and better technology.” [/read]

[Research Project] “Drug overdose deaths have increased in the US. Currently, there have been more than … [read more] 1 million deaths due to drug overdose since 1999. AI technologies have been developed to address a number of issues, such as the automation of mundane tasks and even finding treatment for cancer. Therefore, I propose an AI-mediated warning system of  individuals at risk of overdose at drug-selling stores. Specifically, when a customer buys drugs, the warning system will be given basic information of the customer, such as their sex, history of prescribed drugs, and age. The system will then determine whether the customer is safe or unsafe from drug overdose and follow up with an emergency contact with a close family member or friend of the customer. Two possible AI algorithms have been studied to develop the warning system: neural network and random forest algorithm. Using a training dataset that details individuals and their sex, age, history of prescribed drugs, currently prescribed drug, and state of mortality, the AI model will learn about basic features of an individual that would correspond with their risk of drug overdose. Furthermore, a neural network algorithm  involves the use of weights on inputs that are adjusted accordingly to optimize the accuracy of the AI model. Random forest algorithm involves the creation of decision trees based on random samples of the data. The creation of this AI- mediated warning system will help with the early detection of drug overdose and prevent unnecessary drug overdose deaths.” [/read]

[Research Project] “The research project I did was to aim to leverage artificial intelligence (Al) to address the critical … [read more] issue of drug overdose. The primary focus is on using Al to analyze datasets, such as the Kaggle dataset on accidental drug-related deaths from 2012-2018 and data from the National Institute of Drug Abuse (NIDA). By employing predictive analysis, the research seeks to identify patterns and high-risk areas, enabling proactive intervention. The Kaggle dataset provides a comprehensive view of accidental drug-related deaths, serving as a valuable resource for training predictive models. Classification models, particularly Nearest Neighbors Classification, are proposed to predict drug abuse in specific areas by learning from historical data. This approach involves creating points in proximity to areas with potential risks. Additionally, the research explores the application of Linear Regression Models to forecast trends in drug abuse. This model, relying on input variables, aims to determine whether drug abuse in a given area is likely to increase or decrease over time. By combining these Al-driven approaches, the research project seeks to offer insights that can inform targeted interventions and contribute to mitigating the pervasive problem of drug overdose.” [/read]

[Research Project] “We can use AI to gather online data to enable professionals to predict or determine … [read more] a patient’s mental, physical and emotional distress levels. It does a useful dataset already exist. The data provides data of substance use and mental illness at the national. state, and substance levels, identifies the extent of substance use and mental illness among various subgroups, estimates trends over time, and determines the need for treatment services, it helps. The type of data information need to collected level, pattern, signs and symptoms if use, and consequences of use. Provide a set of tools that can improve discovery and decision making for well-specified questions with abundant, high quality data can be preprocessed feeding ML model. The ML models can be used to accomplished Support Vector Machines, and Neural networks based Deep Learning. The ML algorithm predicts early abuse accompanied by anxiety and depression.” [/read]

[Research Project] “Through this research project, I was able to apply the information I learned about different machine … [read more] learning models into a real life situation. The primary objective of my project is to analyze health records of patients using two different machine learning models in order to detect patterns of indicative substance or alcohol abuse. First, using health records of patients obtained from medical institutions or organizations, patterns would be detected from substance/drug abusers in their medical data. However, because there are many confidentiality issues regarding sharing medical information, it is important to ask or receive approval from the patient to be included in this data. The machine learning model,  Pattern Recognition, which identifies commonalities in data, will be used to detect these patterns. By detecting patterns associated with substance abuse in patients’ medical data, we establish a foundation for subsequent analysis. Next, this data would be applied to the general population or a specific group of people to see if they are abusing alcohol or different substances. The data would run through a different machine learning model, the Decision Tree, which helps to classify if the patient does abuse a substance or drugs or not. Utilizing this Yes or No classification model, a conclusion will be able to be reached regarding if the patient does abuse drugs or substances. If they do, the patient should then be referred to a drug/substance abuse rehab or receive necessary treatment.” [/read]

“I was deeply honored to partake in the GLR-AI program, immersing myself in the exploration of … [read more] ‘The Impact of Artificial Intelligence on Substance Abuse and Alcohol.’ The collaborative efforts with distinguished institutions like Harvard and Stanford offered an unparalleled wealth of expertise. The program’s dedicated focus on AI’s pivotal role in addressing substance abuse, guided by instructors Mirna Kheir Gouda and Aumkar Renavikar, presented a distinctive and enlightening learning experience. During the online classes from December 26th to 30th, I actively applied machine learning models in alignment with the methodology outlined in my project. Beyond enhancing my technical proficiency, the program fostered a profound understanding of the ethical dimensions surrounding AI’s social implications. As the program reached its pinnacle with final presentations on January 6th, I felt well-prepared and confident to contribute substantively to discussions on the intricate intersection of AI and substance abuse. Overall, the GLR-AI program proved to be a truly enriching experience, endowing me with knowledge and insights poised to make a positive impact on substance abuse prevention.” [/read]

[Research Project] “This project aims to tackle the escalating issue of substance abuse, particularly exacerbated … [read more] by the COVID-19 pandemic. The objective is to predict an individual’s susceptibility to drug abuse using machine learning models based on demographic factors. Using Kaggle’s dataset, “Accidental Drug Related Deaths 2012-2018,” the data shows limitations, such as its regional specificity (Connecticut) and the absence of crucial variables like income and family history. The proposed methodology involves preprocessing the dataset, converting categorical variables into numerical values. Two machine learning models – Logistic Regression and K-Nearest Neighbors (KNN), are suggested for prediction. Logistic Regression, a binary classification method, predicts the likelihood of an individual falling into a drug-use category by adjusting weights based on factors like age, sex, and race. KNN, a versatile algorithm, classifies individuals based on the proximity of their demographic features to neighboring data points. The evaluation of model performance will utilize accuracy and F1 score metrics to address false positives and negatives. The ultimate goal is to contribute to the comprehension and prevention of substance abuse by employing demographic predictors to identify potential vulnerabilities. Despite the dataset limitations, this approach provides a valuable step towards understanding and mitigating the complex problem of drug abuse.” [/read]

“My thoughts on the impact of artificial intelligence on substance abuse and alcohol are very positive … [read more] I think that it’s very reliableand good in using ai to impact substance abuse and alcohol. I also feel that its very cool how artificial intelligence can help unite people with the same problems as well as helping them. My experience with that is I have seen people or officers using type of intelligence to test and see what they have. Overall, I think it is a very accurate and smart way to use artificial intelligence for that type of use. I also really like how to the program showed me about different ways and laws of how to use them. The last thing I also think is it’s a great way for people who don’t have access to see to be able to see. What we learned was very new to me and I learned a lot researching about it.” [/read]

[Research Project] “My research project explained two of the ml models I could use to have accomplished my goal … [read more] The two ml models I used were linear regression and logistic regression. I set examples of how they would put into work with the data one of the things i said was u can use face recognition to see if they have used any substance. Another example is i had said we can also use linear regression so the computer could understand it. The way the computer would know with face recognition would be because of the bloodshot eyes and other symptoms of it. Overall, it was two or three slides listing examples of what I learned in the program and ways I think it could be solved through artificial intelligence. It showed an exact example of the way u can use it and how the computer understanded the data. I also explained how its a good use for people who are in need of it from around the world and that don’t have access   to just use artificial intelligence. It also may be useful for when there is no one to see and you can also find out how to fix it as well as learning about it and using for the good of it.” [/read]

“The ‘Impact of Artificial Intelligence on Substance Abuse on Abuse and Alcohol’ GRL Al Class has taught me … [read more] many new things about Al. I am new to Al and did not know much about about python and how to use it but with this class Iwas able to experience a new field of work. Seeing how Al could be used to solve issues such as drug abuse made me realize that machine learning can be very versatile to address different problems.” [/read]

[Research Project] “The intersection of artificial intelligence (Al} and substance abuse represents a complex … [read more] and multifaceted challenge. Al has the potential to significantly impact the prevention, treatment, and understanding of substance abuse and alcohol addiction. On the preventive front, machine learning algorithms can analyze vast datasets to identify patterns and risk factors associated with substance abuse, allowing for targeted intervention strategies. In the realm of treatment, Al-powered tools can personalize therapeutic approaches, tailoring interventions to individual needs and predicting relapse risks.” [/read]

Completing the research program has been a great journey that passed my initial expectations. My experience … [read more] on the program was extremely fulfilling as I learned many new concepts and algorithms in machine learning/AI. Not only did I learn new things through the program. I was also able to hone my problem solving skills. In terms of the class itself, the teachers were exceptionally kind and actually engaged during the lessons, which helped a lot in understanding the material. For example, when I asked questions during the class. I immediately got an answer from my instructor. Another great aspect was that the zoom calls for the lessons did not feel very awkward and dry like many of my online classes I have taken in the past. The collaborative environment fostered engaging discussions and assisted me in gaining more insight on the topic. Overall, the program has helped me realize how machine learning can be used to solve real world problems and it has been a great experience.” [/read]

[Research Project] “The research project I had created for the program involved using machine learning algorithms … [read more] to help solve the problem of people abusing injectable drugs. This issue has caused thousands of deaths across the US and is only getting worse with the rise of individuals taking fentanyl, heroin, and crack cocaine. Following my extensive research on the topic, I proposed the idea of using computer vision AI (more specifically KNN and CNN models), to help combat the problem. The AI would be used in surveillance cameras and such so that law enforcement or emergency medical dispatchers can be notified when someone has taken a drug through a needle, with the AI detecting when that happens. The data needed to execute and train the model would need to be collected and consist of thousands of pictures and videos of people injecting themselves with a needle. This is so that the model can detect when someone is injecting themselves with drugs and become much more accurate in doing so. In order to determine accuracy, the use of a F-1 score can be extremely helpful as it does a great job of balancing the trad- offs between precision and recall metrics. We would not want the model to trigger many false alarms or not detect some people taking the drugs. The two classes the model will be determining from would b a class of not taking drugs and one taking drugs. The input for the model would be frames of video of people and objects and the output would be the people getting classified. Although a CNN model is most likely more fit and accurate than a KNN model because of its ability to further learn and adapt, there may be even better models to use for the idea.” [/read]

It was grateful time to learn about how AI works and many systematic algorithms. When the AI … [read more] first came out to the world, I just thought “oh it’s a just a really advanced search engine like prototype of google” and never thought how we can use AI in many aspects or to solve many social problems. However, through this AI research project, I could learn how AI works specifically, and graft to resolve social problems. Especially for the given topic, “Substance Abuse and Alcohol,” it made me think about specific ways or cases that can help to solve the problem. I liked the curriculum they provided during research program. We could learn basic information before we start and even if we are not familiar to coding or didn’t know how to code, they gave us basic resources that helped to learn basic coding. And later part, we could code follow along the instructors. Again, it was precious experience made me think seriously in aspect of AI and the usability of AI.” [/read]

[Research Project] “My solution for “Drug Abuse and Alcohol” using AI was building 2 different AI model. One was the … [read more] AI system for predictive analysis for High-Risk Individuals. For example, it can be developed based on various factors like social, economic, medical history and behavioral pattern. Using these factors, AI can analysis and predict “most likely” people and we can cope with them previously. The second AI model is for monitoring online platform or uses attachable / wearable technology to identify signs of drug abuse. Through this monitoring the attachable / wearable devices will detect any kind of signals from your body. It would be better if there was a person who went to jail for drug abuse, once they released from prison, make them to wear the device like they tag electronic ankle bracelets to monitor sex offenders. Not only for criminal, but it can also be used as a different form of app using apple watch. It provides timely interventions or alerts to concerned authorities or caregivers. For the data set, using medical records of Anonymous patient data regarding drug prescriptions, or mental health history will be effectful for developing the AI model. And specific patterns indicating drug abuse tendencies from social media, economic and demographic data from government database can be useful too. And there are some needs while preprocessing, data anonymization is necessary to protect privacy. Random Forest Classifier and Neural Networks will be possible ML models, because they are good for handling large datasets with diverse features and Good to recognize patterns in sequence of data which fits best to my planned AI model.” [/read]

“It was grateful time to learn about how AI works and many systematic algorithms. When the AI … [read more] first came out to the world, I just thought “oh it’s a just a really advanced search engine like prototype of google” and never thought how we can use AI in many aspects or to solve many social problems. However, through this AI research project, I could learn how AI works specifically, and graft to resolve social problems. Especially for the given topic, “Substance Abuse and Alcohol,” it made me think about specific ways or cases that can help to solve the problem. I liked the curriculum they provided during research program. We could learn basic information before we start and even if we are not familiar to coding or didn’t know how to code, they gave us basic resources that helped to learn basic coding. And later part, we could code follow along the instructors. Again, it was precious experience made me think seriously in aspect of AI and the usability of AI.” [/read]

[Research Project] “My solution for “Drug Abuse and Alcohol” using AI was building 2 different AI model. One was the … [read more] AI system for predictive analysis for High-Risk Individuals. For example, it can be developed based on various factors like social, economic, medical history and behavioral pattern. Using these factors, AI can analysis and predict “most likely” people and we can cope with them previously. The second AI model is for monitoring online platform or uses attachable / wearable technology to identify signs of drug abuse. Through this monitoring the attachable / wearable devices will detect any kind of signals from your body. It would be better if there was a person who went to jail for drug abuse, once they released from prison, make them to wear the device like they tag electronic ankle bracelets to monitor sex offenders. Not only for criminal, but it can also be used as a different form of app using apple watch. It provides timely interventions or alerts to concerned authorities or caregivers. For the data set, using medical records of Anonymous patient data regarding drug prescriptions, or mental health history will be effectful for developing the AI model. And specific patterns indicating drug abuse tendencies from social media, economic and demographic data from government database can be useful too. And there are some needs while preprocessing, data anonymization is necessary to protect privacy. Random Forest Classifier and Neural Networks will be possible ML models, because they are good for handling large datasets with diverse features and Good to recognize patterns in sequence of data which fits best to my planned AI model.” [/read]

“Being part of the GLR-AI class on ‘The Impact of Artificial Intelligence on Substance Abuse and Alcohol’ … [read more] was an enriching and rewarding experience, as it offered insightful lessons on machine learning and AI foundations. Although I had very little background knowledge of this subject when I joined, the instructors’ expertise in topics like Natural Language Processing and Neural Networks made complex concepts accessible. The interactive lectures and active participation added to the engaging learning environment. The applicational focus, especially through the personal project we created, encouraged our critical thinking and applied what we have learned throughout this class into practices. The personal project emphasized the application of our learned skills, provided a unique and beneficial aspect to the course, making the GLR-AI class an enriching journey into the realm of AI. The Impact of Artificial Intelligence on Substance Abuse and Alcohol module showcased the real-world application of machine learning. Overall, the detailed and structured emails, informative lectures, and the opportunity to work on a self-directed project addressing a widespread issue made the program a valuable and motivating experience.” [/read]

[Research Project] “In my research project, I addressed the need of making a program that would develop a … [read more] solution to preventing drug overdoses, as it is a widespread public health issue, through the algorithms of Artificial Intelligence. By utilizing AI, we have the potential to prevent drug abuse through various methods. One such method involves incorporating psychological data into AI systems, which enables us to predict the risks and contributing factors associated with drug abuse. I researched many solutions, and used machine learning models to make predictions or probability the user is engaging in drug abuse. Neural networks are a deep learning model that identify complex patterns in a variety of data types. It is most used for tasks that have intricate relationships. The patient’s history, multimodal data, including text, images, and structured data were input, and the probability or severity of the drug abuse would be the output. Logistic regression is a binary classification algorithm suitable for predicting the probability of a user engaging in drug abuse based on the input features. You input the patients’ Inputs can include demographic data, health records, and behavioral patterns, and you would be given the probability of drug abuse. Machine learning was effective in developing solutions in the means of preventing drug overdoses.” [/read]

“By taking this class, my perspective and opinion on AI was broadened … [read more] I was able to learn about the coding language, “Python” and not only understood the language itself, but how it could be applied to today’s world. I was also able to learn about how machine learning worked in a variety of ways. For example, for me classification and computer vision were the most interesting as they seemed to be the most applicable for solving the world’s problem with substance and alcohol abuse. It made me realize that AI is able to really help out people and communities that are struggling with these problems. I loved that I was able to hear a wide variety of people’s ideas in the class as many people were from all around the country. Overall, I am very thankful for the opportunity to take this class as it was a great learning experience.” [/read]

[Research Project] “My research project for this class was trying to develop a solution to the ongoing problems … [read more] with drug and alcohol abuse in our country. One idea that I had was utilizing AI to predict early signs of alcohol and drug abuse. A browser plug-in, for example, or desktop program could easily analyze a user’s browser searches, social media activity, and communication patterns-using this data, it could search out for certain keywords and patterns that indicate that the user is beginning to fall into the cycle of addiction. Early detection is especially important for addiction; the sooner the drug abuse is addressed, the more likely it is that an addict will recover. First, we’ll need to compile the necessary data from typical searches and internet activity by drug users as well as analyze financial and language data for indications of abuse. Afterwar, I can see two possible machine-learning models that would be useful for this project: computer vision and classification. Computer vision would allow for Ai to understand the real-world situation as it is important for the AI to have a grasp on the seriousness of the issue. Classification can also input data into predefined classes or labels as I can define if the searches are signs of abuse.” [/read]

“UNSLA opened a new field of interest for me as I was exposed to easily accessible coding programs … [read more] allowing me to explore the concept of AI and its future possible roles in our communities. The GLR-AI program was
special because it connected the lessons to real-life problems as students participated in developing possible solutions for current issues. Through this experience, I became aware of the urgent need for help in terms of drug abuse in the community. Before the program, I had very little knowledge of the severity of drug abuse in our community. Even if I had been aware, I would not think that I would be capable of helping to resolve the issue. The ideas presented in the research projects that students participated in provided original and creative approaches to real-world problems as they utilized their knowledge acquired from the GLR-AI classes. Being part of such a unique program was an honor, and I enjoyed the whole experience very much. I felt that the objectives of the program were executed and planned out very well as I gained lots of knowledge on AI, especially machine learning techniques, and brought my attention to its role in our society.” [/read]

[Research Project] “My project utilized AI to personalize treatments for drug abuse patients … [read more] Drug abuse in America has been growing significantly in the past decade, which calls for help from medical professionals to help overdose patients. For a successful treatment, it must be highly personalized to the patient’s needs. Machine learning AI can easily keep track of millions of patients and can personalize treatments to a patient far better than most doctors due to its ability to learn to predict data. The Data would be collected through various forms of health monitors such as electronic watches. These AI monitors will specifically monitor drug abuse patients, storing and tracking their health patterns. Two machine-learning models are used for this solution, which are linear regression and the decision tree model. Linear regression is used to predict numerical values of health levels from the stored data. Through this model, the AI machine will determine the relationships between certain factors to increased/decreased use of drugs. The classification model or decision tree model is used to determine the optimal treatment for an individual from the data in the linear regression model. For instance, if the patient frequently uses drugs in the evening, the Fitbit will detect changes in the individual’s health levels, which through the linear regression model can be predicted and concluded by the machine that the individual is likely to take drugs in the evening. This information will be used in the decision tree model as it determines the severity of the individual’s addiction.” [/read]