This project demonstrates a unique approach to weather forecasting by integrating IoT-enabled sensors with Machine-Learning techniques. The system is designed to collect real-time environmental data such as temperature, humidity, and pressure using a BME680 sensor and ESP32 microcontroller connected via a wifi. This data is uploaded to a Google Firebase cloud server, where it's data is used in preprocessing and noise filtering using a Kalman filter, ensuring clean and reliable input for machine learning models.
The filtered data is used to train machine learning models, including LightGBM, which achieved the highest prediction accuracy among all the models with an R² score of 0.997. The processed data and predictions are displayed on a dynamic, user-friendly web application hosted on Google Firebase, allowing users to view live updates without manual refresh.

The system includes advanced preprocessing, such as normalizing data, removing noisy columns, and analyzing variations using timestamp-based segmentation. The Kalman filter enhances data accuracy by removing noise and outliers, while LightGBM leverages its efficiency and accuracy to deliver robust forecasts.