SMART WEATHER PREDICTION FROM SENSOR DATA USING MACHINE LEARNING

Abstract

In this study, we implemented a weather prediction model using multiple classification algorithms like logistic regression, KNN, SVM, Decision Tree, Naive Bayes, and XGBoost. Notably, the Random Forest classification model emerged as the most effective for forecasting diverse weather parameters like drizzle, rain, sun, snow, and fog. Leveraging historical data, machine learning enhances weather forecasting accuracy by identifying patterns and handling complex relationships, integrating various data sources like satellite imagery. The model, utilizing a dataset with features such as temperature, humidity, wind speed, and atmospheric pressure, underwent preprocessing for missing values and feature normalization. The Random Forest algorithm demonstrated an 86% accuracy, validated by the confusion matrix analysis during training and evaluation. This study underscores the Random Forest algorithm's efficacy in multiclass weather prediction, emphasizing its potential to revolutionize forecasting accuracy and planning capabilities, outperforming existing strategies in precision and computational efficiency.