Machine learning classification of rainfall forecasts using Austin weather data

Ting Tin Tin, Enoch Hii Chen Sheng, Loo Seng Xian, Lee Pei Yee, Yeap Sheng Kit

Abstract

The paper examines the machine learning classification of rainfall forecasts using Austin weather data. Rain is a natural phenomenon that is essential for the Earth's water cycle. Rain brings benefits to daily lives and also causes disasters, such as floods, which will endanger lives in addition to causing great losses. Due to this, many methods have been studied and experimented with to find a solution to predict rainfall and prevent tragedies from happening. In this research, the Austin weather dataset is applied to make predictions of rainfall through the implementation of machine learning models. The models used to predict rainfall based on the data set were Extreme Gradient Boosting, Support Vector Machine, Long Short-Term Memory, and Random Forest models. 21 variables with 1319 records were present in the dataset, but the variables used for the modelling were 18 variables from the original data, and 1 variable, “Precipitation Sum,” was converted to the variable “Precipitation Range,” which contained the classes “no rain,” “small rain,” “moderate rain,” and “heavy rain” based on specific value ranges. After training and predicting the data on the models, it was shown that Extreme Gradient Boosting gave the best results of 85.17% accuracy, 83.19% F1 score, 85.17% recall score, and 82.14% precision score, and was able to give predictions on all 4 classes of rainfall. This study and the way to implement machine learning models for rainfall prediction have the potential to provide new insights and methodologies for future studies and pave the way for finding a high-accuracy rainfall prediction method to avoid disaster.

Authors

Ting Tin Tin
tintin.ting@newinti.edu.my (Primary Contact)
Enoch Hii Chen Sheng
Loo Seng Xian
Lee Pei Yee
Yeap Sheng Kit

Article Details

No Related Submission Found