Neural Network Architecture
Sample Run Snapshot
- How important is my GPA to me, as a student?
- How important are your students’ grades to you, as faculty/administration?
- How often do we struggle selecting the best course to improve our GPA?
- How often do we drop certain courses to avoid failure despite no reimbursement?
These are common problems for both faculty and students at the beginning of each new semester. So we thought of a solution to ease these problems, the result of which is MyNextA+.
What it does
Our model predicts grades for students in the senior-level courses based on their performance in freshman, sophomore and junior levels and majors.
How we built it
We received a dataset of students in the following format in csv:
Student_ID | Course | Level | Grade | Major
We processed it to obtain two dataframes to be fed as input and actual output to our neural network:
Student_ID| Course 1 A | Course 2 B| .... | Major 1 | Major 2 ... | Major N
Student_ID| Senior Course 1 A | Senior Course 2 B
Challenges we ran into
- Extracting courses taken only during a single level
- Lack of information about instructors of particular courses
Accomplishments that we're proud of
- We achieved over 85% accuracy
What we learned
- Importance of Plan B. This was our second plan in case we fail in our mobile app.
- Learnt Scikit Learn Library in Python
What's next for MyNextA+
- Incorporate instructor details as input features to improve accuracy
- Filter out similar courses and build smaller neural networks for specialized majors