Applying Machine Learning Technology for Weather Forecasting: A Case Study of the Logistic Regression Model

This study is focused on improving the dependability and precision of weather forecasting by employing the capabilities of Artificial Intelligence. Specifically, this study utilizes Logistic Regression and Machine Learning techniques to forecast weather, demonstrating the potential in optimizing weather-related activities and disaster management strategies. The study relies on comprehensive weather data observed over several years, sourced from Kaggle, and handles missing data and outliers during its pre-processing stages. The primary machine learning tool applied is Logistic Regression, followed by a stepwise feature selection to identify influential features for accurate weather prediction. The workflow also involves data collection, pre-processing, model building, training, and testing, with provisions for handling both numeric and categorical features along with imputations. The accuracy, precision, and recall of the prediction module are tested using appropriate statistical tools. The Logistic Regression model, upon implementation, demonstrated considerable accuracy, with an ability to predict rainy days and non-rainy days efficiently. An analytical approach was used to examine the model's sensitivity towards the removal of each feature, thereby ascertaining the relative importance of each. Critical predictors like 'Rainfall', 'Pressure9am', and 'WindGustSpeed' exhibited significant effects on the probability of rain. Overall, the use of Logistic Regression and Machine Learning techniques notably improved rain prediction, offering potential for further advancements in the field of weather forecasting.