Improving Backpropagation Performance for Air Humidity Prediction Using Genetic Algorithm-Based Feature Subset Selection

Air humidity is a critical parameter in weather and climate prediction, particularly in anticipating extreme weather conditions such as storms, floods, and heavy rainfall in tropical regions. As such, air humidity prediction technology is essential for identifying potential changes related to humidity levels. This study aimed to develop an air humidity prediction model by leveraging a Backpropagation algorithm optimized through Genetic Algorithm-based feature selection. The goal was to improve prediction accuracy and reduce model complexity by selecting the most relevant features from the available dataset. The Genetic Algorithm successfully identified three key features out of the 14 available, which consisted of Atmospheric Pressure (Po), Total Cloud Cover (N), and Horizontal Visibility (VV). The air humidity prediction model was then constructed using the Backpropagation method based on the selected feature subset. Testing of this model demonstrated significant results, with a Mean Squared Error of 290.617 and a Mean Absolute Percentage Error of 19.03%, indicating good model performance. Moreover, this integration also reduced computation time, demonstrating the model's efficiency in processing large datasets. This research highlights that integrating a Genetic Algorithm for feature selection can significantly enhance the performance of Backpropagation models in air humidity prediction, offering valuable insights for future applications in weather forecasting and climate adaptation strategies.