Demand Forecasting in Retail for Diabetes Medicine

This paper focuses on forecasting demand for diabetes medicines over a short period of time. Retailers face the problem of stock out or overstock due to improper management or estimation skills. So, to help retailers overcome these loss causing problems, our system predicts the diabetes medicine requirements based on the previous sales data. This proposed work has implemented mainly two algorithms, Auto Regressive Integrated Moving Average and Long Short Term Memory shortly known as ARIMA and LSTM respectively. Among these two algorithms, used for demand forecasting, an appropriate one is to be selected which is capable of providing a high estimation accuracy. Our experimental results measured with the help of real time dataset shows that during the training phase, LSTM demonstrates superior performance. However, in predicting the values, ARIMA outperforms LSTM.