Inspiration
There has been a rise in the number of school shootings, and the most unfortunate part is that the warnings signs were there. Most frequently, school shooters make a post on social media about their thoughts and intentions, and they go unnoticed and unreported with disastrous consequences. It is our hope that, by monitoring users that make posts with alarming language, we can help students get through mental illness and prevent another tragedy.
What it does
Detects certain keywords, phrases, and images to start catching early signs of stress/depression that may lead to a more severe situation. Convolutional neural networks are used for the detection of guns in images. We also used a Levenshtein algorithm was text comparison to keywords that may indicate mental problems.
How we built it
We conducted research on depression and suicidal behavior to build a data set of phrases and categories that could be compared to a social media post of a person. Python was used to determine text similarity (higher similarity = more risk of depression). We also trained our model for gun detection in Python. We then set up the communication between the back-end and front-end using ajax technology. The front-end contains user-friendly animations, useful statistics, and a way for students and administrators to use the service. Our teamwork was very effective. Each person had an area of specialization (front-end, back-end, machine-learning, or research). Thus, we worked quickly and efficiently.
Challenges we ran into
We had some animation problems in the beginning because we were unfamiliar with the library we were using. Also, our neural network model took a very long time to train because we did not have powerful computers and had a lot of data. Of course, we also encountered the typical problems like syntax errors and logic errors. However, as a team, we dealt with those errors in a very effective manner.
Accomplishments that we're proud of
Every person has something they are proud of in this project. Nick is proud of his text evaluation classes and Twitter API integration. Amil is proud of his high accuracy on his CNN. Jesus is proud of the communication between Python and AJAX. Ahmed is proud of his animations. Cayden is proud of the great number of research he did. Ji Yoon is proud of her drawing, that are working with our animations, and her marketing strategy. Charles (mentor) helped us with keyword generation.
What we learned
We learned how to use an animation library, the Levenshtein algorithm, and convolutional neural networks. Overall, we greatly improved our technical skills.
What's next for Forecast
In terms of technical development, we are planning on doing more with our machine learning, including adding image recognition to be able to identify weapons in image posts as well as obtaining a better, more comprehensive data set to train from. To work on marketing, we plan on making our target audience schools, specifically teachers, administration, and counselors, and parents. We will implement a three ponged approach that goes beyond just contacting the influential figures of a given school, taking advantage of products that our targets commonly purchase, committee meetings, and advocates for mental health awareness, especially in students. Products that our targets use, such as school supplies and textbooks, are easy ways to advertise. Asking online school supplies catalogs and publishing companies to put a small informational advertisement about mental health and our solution on packaging and/or on their website, both parties can benefit. The company is able to show their customers that they are advocates of student mental health awareness, and we are able to provide an option to help schools deal with mental health issues. Schools often have meetings to discuss problems and possible solutions, which would be great ways to pitch our service to the school. Not only in the school's board meetings, but PTA, or the parent association equivalent, meetings can also bring awareness about the issue and our solution to parents concerned about their children's well-being. Counselors also have meetings. For example, the ISCA, Illinois School Counselor Association, has an annual conference, a good opportunity for us to pitch our service. There are plenty of mental health advocates that would be more than willing to spread word about our service and provide insight and help us with recommending steps to take to help school administrations with their students' mental health.
Log in or sign up for Devpost to join the conversation.