Application of Mutual Information Estimation Methods for Feature Selection in Regression Problems of Aviation Traffic Forecasting

In many machine learning applications, high-dimensional data complicates processing, making feature selection critical. This work aims to review and compare existing feature selection methods for regression tasks using real data. The study focuses on using mutual information-based methods to address the impact of economic indicators on passenger air transport volume. We discuss the principles, advantages, and limitations of these methods and provide practical examples. Comparative studies were conducted using regression models like support vector machines, multilayer perceptron, decision trees, and k-nearest neighbors, highlighting the benefits and drawbacks of mutual information-based feature selection methods.