AI Competition. Time Series Analysis - Power Generation Forecast

Participated in an AI Hackathon to develop a time series regression model for predicting wind turbine power generation. Preprocessed the data and engineered features like mean, rolling averages, and cyclical fields. Experimented with Random Forest, Linear Regression, LSTM, XGBoost, and CNN models. Fine-tuned hyperparameters using Optuna and Grid Search. Leveraged PyTorch to learn the framework. Most accurate model was an ensembled XGBoost and Random Forest model which achieved an RMSE of 0.185, securing 4th place out of 50 participants.

https://machinehack.com/hackathons/wind_turbine_power_generation_forecasting/leaderboard

Learning Objectives : Time Series Analysis, Pytorch, Data Processing, Feature Engineering, Model choice, Ensembled model, Hyperparmter Tuning Optuna & Gridsearch, Prediction Forecasting