Predicting Drug Sale Quantity Using Machine Learning
Warayut Saena, Vasin Suttichaya · 2019
Medication is one of the essential parts of a patient's treatment. Therefore, it is important to have good medication storage administration in order to have effective medication storage. This study aimed to find a proper model used for the prediction of medication purchase amount by using machine learning to analyze medication purchasing amounts in the form of time series. In this research, the first 10 medicines in AV group were chosen. Then, Multilayer Perceptron (MLP), Long Shot-Term Memory (LSTM), and 1D Convolutional neural network with LSTM models were used together with Rolling Windows which were used to predict the purchase amount of each model. The periods of prediction were at 1 month, 3 months, and 6 months. The efficacy of each model was compared using their errors. CNN-LSTM model produces the better forecasting results. The result also shows that 1-month forecasting period is suitable for medicines that specific to disease. The 3-month forecasting period is suitable for commonly used medicines. The 6-month forecasting period is suitable for the medicines for chronic diseases.