Perbandingan Teknik Prediksi Pemakaian Obat Menggunakan Algoritma Simple Linear Regression dan Support Vector Regression

Public Health Centers (Puskesmas) had a crucial role in furnishing society essential healthcare services and medication management. To preempt errors in stock management, a predictive approach is employed. This prediction methodology involves comparing Data Mining techniques utilizing the Simple Linear Regression algorithm and Machine Learning methodologies harnessing the Support Vector Regression algorithm. This research uses Paracetamol 500 mg and Cetirizine drug data from January 2020 to June 2023. The selection of these algorithms is motivated by the continuous nature of the data variables and their temporal span, spanning 42 months (period). The core aim of this study is to evaluate the magnitude of predictive errors using the Mean Absolute Percentage Error (MAPE) methodology. Implementing these methods was effectuated through the programming language Python with an 80%:20% partitioning of training and testing data. Drawing from experimental endeavors conducted concerning Paracetamol 500 mg, the utilization of the Simple Linear Regression algorithm, yields a MAPE score of 20.85%, categorized as 'Moderate,' whereas the application of the Support Vector Regression algorithm generates a MAPE of 18.39%, classified as 'Good.' Otherwise, experimentation on Cetirizine employing the Simple Linear Regression algorithm, employing an identical division of training and testing data, results in a MAPE of 18.39%, also classified as 'Good.' Meanwhile, resorting to the Support Vector Regression algorithm leads to a MAPE of 17.14%, falling under the 'Good' category. Based on the MAPE obtained, the Support Vector Regression algorithm has better prediction results than the Simple Linear Regression algorithm