Navigate to Projects User Profile

Exam Score Prediction

Description

This neural network predicts a students exam score, based on features such as their sleep hours, previous exam scores, peer influence, and more.

Problem

Me and my friends wanted to know what exam score would be achievable based on certain factors about us

Plan/Solution

Materials Used

Visual Studio Code

Data

  • Python
  • Extensions
    • Pandas
    • Tensorflow
      • keras
      • Dense layers
    • Matplot lib
    • Sklearn test train split
  • CSV Data of student features and final exam scores to train and validate
  • Visual Studio Code for IDE

How It Works

  • The model learns based off of these features: and the prediction target is the final exam score
  • It is fitted upon training data, split from the student performance
  • The model then predicts an exam score based on what it has learned
  • It then compares its predictions to the other split of the data: the validation data
  • After this, the model continues to adjust its weights in hope of reducing the loss to its minimum

Building Process

  • After each iteration of code I had, I used the graph of validation training loss, as well as the minimum training loss at the end, to alter my code
  • This process continued until I reached a reasonable error of 0.47 (current model)

Results

I was able to give a decent prediction on my final exam score, based on the given factors.

Challenges/Lessons Learned

My model was overfitting, learning unwanted noise.

  • I implemented Dropout layers, to reduce the noise learnt