<![CDATA[The Loop Prophet method is a time series forecasting approach that utilizes the Facebook Prophet model iteratively (looping) to automatically process multiple variables or keywords. This study aims to predict the weekly search trend on Google for the leading causes of death in Indonesia, taking into account trend, yearly seasonality, and weekly seasonality components, as well as to evaluate model performance using error metrics such as NMAE, NRMSE, and MAPE. The dataset consists of weekly Google Trends search data for selected diseases during the period from September 2020 to July 2025. The results indicate that the Loop Prophet model successfully captures both trend and seasonal patterns. Based on the evaluation criteria, models categorized as “very good” were obtained for the keywords “Stroke”, “Diabetes”, and “Diare”. The “good” category was obtained for “Serangan Jantung”, “Sirosis Hati”, “Penyakit Paru”, “COPD”, dan “Neonatal”. The keyword of “Tuberkulosis” was categorized as “good enough”. Meanwhile, the “poor” category was found for “Kanker Paru-Paru” dan “Pneumonia”, which tend to have fluctuating patterns influenced by incidental events. These findings demonstrate that the Loop Prophet method is effective for time series analysis with complex seasonal patterns, although its performance decreases for diseases with highly irregular search trends.]]>
Copyrights © 2026
