Ethiopian Health Commodity Demand Prediction

Abstract

A pharmaceutical agency’s ability to forecast demand accurately is requisite to identify, rationalize and improve results. The optimization of order quantity, stock level, or delivery schedule depends on forecasting accurately, demand at the store level. Forecasting is important in the perspective of the pharmaceutical areas, which commonly requires efficient Supply Chain Management. The main objective of study was to forecast demand by using EPSA consumption data. EPSA consumption data was extracted from HCMIS database from the year 2016 to 2019. After the data is converted into model understandable format, comparative study is done on three algorithms (LSTM, Multi-Layer Perceptron and Auto ARIMA). The best performing algorithm is used to forecast the future. By using MSE metric, we conclude that Auto ARIMA gives the best prediction performance from the three algorithms. It also has a closer result comparing with the actual data. The MSE results for Auto ARIMA in the selected four items are Atropine Sulphate - 1mgml - Injection (Sulphate) is 329, Tetanus Antitoxin (Human) Equine - 1500 Units – Injection is 3,747, Insulin Isophane Human - 100IUml in 10ml Vial – Injection (Suspension) is 36,290 and Dextrose - 40% in 20ml - Intravenous Infusion is 173,979 respectively. Predicting demands accurately also helps to accommodate time, cost, flexibility, space, efficient inventory management, wastage reduction and minimize product expiry properly.