Devansh G.

Stock Price Prediction Using Market Indicators

• Designed a hybrid stock prediction pipeline for Indian equities using daily and 5-minute OHLCV data from Yahoo Finance. • Built incremental data collection, cleaning, validation and Parquet-based storage workflows with metadata tracking. • Segmented stocks into 6 datasets by market cap and liquidity, then into 24 volatility–trend market regimes. • Engineered 300+ candidate features across price, momentum, volatility, volume, liquidity and market structure. • Developed multi-stage feature selection (Spearman clustering, Mutual Information, XGBoost, SHAP) with chronological, leakage-aware validation.

AutoML function

Takes in a pandas dataframe, performs training of several models on it after train test splitting (takes in the user's grid search parameters, which models to train, which scoring Metric to prioritise, regression or classification) and gives the scores of various models trained as output