MiniViz May - Week 2 | Machine learning test data

This report summarizes a machine learning project developed in Python using historical hourly time-series data from 1950 to today. The objective was to build and evaluate a multi-output regression model capable of predicting the next 36 hourly values simultaneously based on historical observations and engineered time-series features.

The dataset was split chronologically into training and test periods to avoid data leakage. The resulting metadata, prediction outputs, and forecast error metrics were imported into Power BI for interactive analysis and visualization.

The report focuses on model behavior, prediction accuracy, and forecast uncertainty across different prediction horizons.