AI research
LSTM Forecast
Stock price prediction with LSTM networks and an honest look at the limits.

01 / Challenge
Evaluate whether LSTM networks can forecast financial time series better than simple baselines, and document where they fail.
02 / What I built
A research pipeline with engineered features from historical prices and technical indicators, tuned LSTM models and baseline comparisons.
03 / How it works
Sliding-window datasets, feature scaling, hyperparameter search and walk-forward validation. Metrics: RMSE, MAE and R² against naive and linear baselines.
04 / What is there now
Notebook and report with charts of predicted vs actual prices, loss curves and error analysis.
05 / Result
Documented findings on overfitting, volatility impact and why good-looking charts can mislead (12/2025 – 01/2026).