Overview
As part of my bachelor’s capstone, I co-authored and published *“Feasibility of the Neural Network Model for Stock Price Prediction”* in the Journal of Industrial Information Technology and Application (2024).
We trained a 1D CNN on three years of Tesla and Disney data, with leakage-safe feature engineering and walk-forward validation. Results showed CNNs can rival recurrent models in stability, proving that data discipline matters more than complexity.
Outcomes
Validated CNNs as a feasible alternative to LSTMs for short-term forecasting.
Reinforced the importance of rigorous validation in financial modeling.