Grade Predictor

Inspiration

We wanted to provide college students just like us a place to document their studying habits and get an accurate estimate of their academic standing. Our goal was to find a way to implement machine learning algorithms and data from student performance vs studying habits, and find a correlation between such to create a representation model of how a student may preform in the class.

What it does

The model will take a students study time, grades from multiple periods, past failures, and absences and predict a possible score for your overall final performance in the class. The web app is honesty based, meaning that it is up to the user to provide accurate information based on their current performance to gain a realistic outlook on their final performance.

How we built it

We decided to use a machine learning model incorporating python scripts within Google Collab. We decided to enclose html and css code within python cells. This was created through a separate app.py which was then designed to be published using Streamlit, and to incorporate a gemini api that gives you advise. We also used panda, which is a python programming library that is made for data manipulation and analysis. The data we collected to create prediction models was from a paper published by P. Cortez, A. M. G. Silva. 2008 titled "Using data mining to predict secondary school student performance."

Challenges we ran into

We faced some pretty frustrating challenges. One significant problem was moving away from using a local tunnel in Collab to an actual web app using Streamlit. Another issue was managing the css and translating it to Streamlit to make the web app easy on the eyes and actually look good. It was a bunch of trial and error for that one and we eventually agreed on a green theme because who doesn't love green? Some challenges when developing the model included the data set including individuals who dropped out, as well as some of the target audience being unreached.

Accomplishments that we're proud of

We are honestly proud of how successful the model was at generating predictions. It has a wide range of outcomes and is extremely flexible. One thing we are definitely proud of is managing all of the bugs we came across. It was definitely a challenge but through team work, we managed to fight off some of the bugs that plagued us the most. We are also proud of coming up with simple UI design that is interactive and effective.

What we learned

We learned effective machine learning and data analysis using adaptive prediction models. We learned AI integration, and how we can use that to search and generate effective studying tips for success. Additionally, we were able to learn UI design to create an easy and effective web layout that is both recognizable and simple.

What's next for Grade Predictor

Our mission is to help students like us become better prepared for upcoming assessments and checkpoints in their learning. We want to be able to expand this further by improving our grade weighing scale that is also as adaptable. We also want to implement popular studying techniques within the app such as flashcards or a Feynman Technique system where you are challenged to explain topics in the most simple way possible while being asked questions based on your explanations.