Time Series Forecasting for Artist Investment Decisions

This Hackathon Challenge aims to evaluate whether an artist is likely to generate enough streams to justify an investment, helping the company make informed decisions and drive revenue. I used TimeGPT on time series data to learn from historical patterns, predict future tendencies, and provide confidence intervals for more reliable, data-driven evaluations.

Human Activity Recognition

This is a video classification-based project where we used several techniques and compared them: Simple CNN, CNN + LSTM, Pose Detection, 3D CNN, and the Slow Fusion technique.