ClustClass: Blending traditional meteorology with machine learning for rainfall forecasting

This paper explores integrating machine learning techniques in rainfall forecasting to enhance accuracy and reliability in predicting meteorological events. The study proposes a novel hybrid model that combines K-means clustering, Principal Component Analysis (PCA) for dimensionality reduction, and Decision Tree classification. The research utilizes a comprehensive dataset of historical weather data from Tehran, Iran, from ۲۰۱۴ to ۲۰۲۴. Through a systematic approach, the model preprocesses the data to ensure quality and then applies PCA to retain critical information while reducing dimensionality. Two instances of K-means clustering categorize the data, which is subsequently classified using a Decision Tree algorithm. The results demonstrate exceptional performance metrics, achieving perfect scores in accuracy, precision, recall, and F۱-score, reflecting the model's robust capability to capture complex rainfall patterns. Moreover, the findings underscore the importance of accurate rainfall predictions in applications such as disaster management, water resource allocation, and agricultural planning. The proposed model showcases the potential of blending traditional meteorological methods with advanced computational techniques, presenting valuable insights for future rainfall forecasting endeavors.

کلیدواژه ها:

rainfall forecasting ، machine learning ، K-means clustering ، decision tree ، PCA ، meteorology