Comparative Accuracy of Data Mining Models for Predicting Extreme Weather Events in West Sumatra

Abstract

Weather factors have a vital role in human activities, especially extreme weather phenomena. Extreme weather can result in potential hydrometeorological disasters that cause loss of life and property. Climate change is also contributing to the higher frequency of extreme weather events. For this reason, research related to predicting extreme weather, especially very heavy rain, is needed to anticipate its impact. Research related to the prediction of extreme weather events is currently still being carried out using various models. By utilizing aerial observation data from Radiosonde (RASON) and daily rainfall data at the Minangkabau Meteorological Station Padang
Pariaman, West Sumatra, extreme weather prediction modeling was carried out with the criteria of heavy rain events having rainfall intensity above 50 mm/day or 50 mm/day. 24 hours. From the data mining prediction model that has been carried out using the Support Vector Machine (SVM) Model, in this case, the Support Vector Regression (SVR), the Mean Squared Error (MSE) value is 502.88, and the R2 (Coefficient of Determination) score is 0.09. For the Artificial Neural Network (ANN) model, the Mean Squared Error (MSE) value was 590.03, and the R2 (Coefficient of Determination) score was -0.73 with an accuracy value of only 0.11 and a loss model value of
590. Meanwhile, for the data mining classification model using the Decision Tree Model, the value obtained The model accuracy was 0.47, and the Naïve Bayes (NB) model obtained a model accuracy value of 0.34. From the results of this comparison, it was found that the prediction model using the Decision Tree Model was more accurate in predicting extreme rain events in the West Sumatra region.