This project predicts the next day’s Nifty closing price using a Multi-Layer Perceptron (MLP) model trained on 1 year of Indian stock market data (Nifty200).
We processed the data, normalized it, created rolling sequences of 30 days, and trained a deep neural network to forecast the following day’s value.
- Preprocessing: Cleans missing values and normalizes stock prices
- Sequence Modeling: Uses past 30 days to predict the next day
- MLP Neural Network: 4 hidden layers with ReLU activation
- Evaluation Metrics: MAE, RMSE, R², MAPE, Accuracy
- Visualization:
- Loss curves across training epochs
- Actual vs Predicted next-day prices on real examples
Model Performance:
MAE : 69.56 points
RMSE : 96.57 points
R² : 0.9687
MAPE : 0.52%
Accuracy ≈ 99.48%
