A Chess AI With Poor Eyesight

Oh, no! The computer lost its glasses! This chess AI can only recognize the location and color of pieces on board, and was trained with a deep-learning model on a database of millions of games played by real people to be able to guess the correct board state when it makes a move. See the Github (link) or just the report (link).

Keywords: data mining, feature engineering, classification, class imbalance, deep learning, PyTorch, parameter tuning, ensemble modeling, model deployment, dashboard

Multi-Location/Multi-Product Demand Forecasting

Forecasting demand of a variety of products for branches of a chain store scattered across multiple cities. I evaluated and compared multiple model structures, informed by analysis of the data. See the Github (link) or just the report (link).

Keywords: forecasting, regression, time series, auto-regressive (AR) models, XGBoost, ensemble modeling, parameter tuning, feature selection, table joins, dashboard

Predicting Chess Game Winner During the Middlegame

Predicting the winner of a chess game after turn 20 given the board state and history of the game to that point, trained on millions of games with 25+ turns played. See the Github (link) or just the report (link).

Keywords: data mining, feature engineering, classification, logistic regression, XGBoost

Forecasting Wind Power Generation

A suite of forecasting models predicting wind-power generation from a single turbine. See the Github (link) or just the report (link).

Keywords: data cleaning, feature selection, forecasting, linear regression, SARIMA, XGBoost