Click on the image above to watch a demo on YouTube!

Overview:

Developed a machine learning tool using scikit-learn’s ridge regression to provide hyper-local weather predictions, accurately forecasting next-day temperatures within a 2-degree margin.
Trained the model on historical weather data from local weather station datasets, using predictors like temp_max, temp_min, and the previous day’s precipitation to generate precise predictions.

Problem:

Often, local weather stations cannot report accurate forecasts relative to your specific location.
However, many weather stations have public repositories of historical weather statistics, which may be closer to your specific location.

Solution:

Utilizing publicly available weather datasets (ie National Oceanic and Atmospheric Administration) to provide accurate machine learning location-based weather forecasts.

Future Works:

Ideally, I'd like to be able to predict accurately more than one day in advance while maintaining a 2-degree margin of error.
Auto update max_temp, min_temp and precip values using APIs