A Study on the performance of Four Regression Models in Predicting Weather Temperature Based on Python

Faculty of Science and Engineering, University of Nottingham Ningbo, Ningbo, China

Corresponding author: scytl7@nottingham.edu.cn

Abstract

For industries like agriculture and disaster management, weather forecasting is essential. This study assesses how well four regression models—linear regression, random forest regression, support vector regression (SVR), and K-Nearest Neighbors (KNN)—predict weather temperatures using a dataset from England. Standardizing and expanding features were part of the data preprocessing process to capture non-linear interactions. Performance metrics were used to evaluate the models' predictive capacity. With the highest R2 value and the lowest error metrics, Random Forest Regression fared better than the other models, suggesting higher predictive accuracy, according to the data. KNN exhibited greater sensitivity to local fluctuations compared to SVR, which performed slightly better overall. linear Regression was the least effective, struggling with non-linear data and exhibiting higher error metrics. This study offers a thorough comparison of weather prediction regression models, emphasizing the performance of the Random Forest regression.