MODEL HYBRID MARS ARIMA FOR TRIBAL-BASED MALARIA PREDICTION IN TANAH BUMBU DISTRICT, SOUTH KALIMANTAN

Abstract

Tanah Bumbu Regency has the highest rate of malaria in South Kalimantan Province. Due to the non-linear fluctuations in malaria cases by ethnicity, a hybrid model combining Autoregressive Integrated Moving Average (MARS ARIMA) and Multivariate Adaptive Regression Splines was proposed for time series forecasting. The purpose of this study is to use the MARS ARIMA hybrid model to predict malaria cases by ethnicity in Tanah Bumbu Regency. The findings demonstrate that the best inputs for MARS modeling are significant lags found using ACF and PACF. The hybrid MARS ARIMA model performs better than standalone ARIMA or MARS models, according to predictions. Key findings show that the number of patients over 35 during the preceding two periods influences increases in malaria cases for the Banjar ethnic group. Cases exceeding 13 in two prior periods and 19 in one prior period are associated with increases for the Javanese group. Cases of more than two or fewer than two in the preceding two periods and more than eleven in one preceding period have an impact on increases among the Bugis. Prior cases below 26 have an impact on Banjar case declines, whereas prior cases below 13 and above 3 have a significant impact on Javanese case declines. This study demonstrates how well the MARS ARIMA hybrid model predicts malaria cases according to ethnicity.