AlcoDetector

Description

By Matthew Fan, Jeremy Yu, and Bill Ma:

Our app, AlcoDetector, aims to protect students from the dangers of underage alcohol consumption by utilizing machine learning. With the help of an artificial neural network model we created and trained, our app diagnoses any student’s risk of being exposed to and consuming alcohol using key information about the student. The user, which can be a school or institution, the student’s friends or family, or even the student themself, can take advantage of our app to learn their risk and take the necessary action to ensure safety from alcohol. Our app’s user interface is convenient and easy to understand; after filling out a survey, the app shows the user the alcohol risk factor on the scale of 1-5 (with 1 being no risk of drinking and 5 being hazardous risk of drinking). High risk levels also identify not only those who are likely drinking but also those who are likely to be exposed to alcohol. Furthermore, schools and institutions can use the mass input feature we implemented to diagnose risk for hundreds of students at a time; this functionality allows institutions to identify high risk individuals and to intervene and protect their students.

Student alcohol consumption is a widespread and growing issue all throughout the United States. All three of us have witnessed the grave consequences of alcohol consumption firsthand. After seeing several of our friends struggle with poor academic performance, negative emotions, and low energy levels, we knew that something had to be done. We noticed, however, that these students had practically no help with and intention of escaping their alcohol problem, as most of the adults in their life had no idea this was happening. We thought, however, that if each student’s parents could at least know the risk their child had of alcohol consumption, then addiction and habitual drinking could be more easily prevented.

As peer pressure, among other factors, led to more and more of our friends drinking, we realized that millions of students across the nation were struggling with the same problem but lacking the help they needed. So, we decided to take action by applying our knowledge in machine learning and app development and ultimately creating this app to help our peers.

Going forward, a 2.0 version of our app would likely include a feature that recommends a course of action for the user to take. For example, if our app diagnosed the student with a hazardous risk factor of 5, then the app would recommend immediate intervention such as speaking to the student or contacting the student’s parents if it was an institution. If it was the student themself using our app, a better solution could be to contact another adult or organization if they feel uncomfortable speaking about the issue with their parents. On the other hand, if our app diagnosed the student with a low risk factor of 1, then the app would tell the user that further preventative action is unnecessary while still advising against drinking.

Another possible area of improvement is our app’s accuracy in diagnosing risk. Our app employs machine learning through an artificial neural network model we developed and trained. Out of hundreds of test cases, our app currently outputs the correct risk factor with slightly over 90 percent accuracy. A perk of utilizing machine learning, however, is that as time passes, we can gradually improve our model’s accuracy by using anonymously collected user data if the individual allows it.