Demand Forecasting in Pharmaceutical Industry Under Covid-19 Pandemic Conditions by Machine Learning and Time Series Analysis

As pharmaceutical products carry vital importance for society, demand forecasting of pharmaceuticals is much more critical. A well-designed demand forecasting and planning can prevent pharmaceutical companies from stock-out and high disposal costs of products. However, there is a limited number of studies about demand forecasting in the pharmaceutical industry, especially in pandemic conditions. This article aims to examine this under-researched area and understand the factors that affect the demand for pharmaceuticals significantly in pandemics, and hence perform an accurate demand forecasting. In light of the literature review, the factors affecting the demand for the pharmaceutical are historical sales, price, promotion factors, campaigns, currency rates, market share, and seasonal or epidemic diseases. Since the chosen pharmaceutical product is used in enteric diseases treatments and lockdowns prevent access to public places, the Covid-19 pandemic is thought to be a factor affecting the sales of the selected product. The forecasting methods of Holt-Winter exponential smoothing, multiple linear regression, Artificial Neural Network, and XGBoost were applied. According to the results, XGBoost was determined as the method that gave the best forecasts, and significant factors affecting the demand were determined. This study is the first one in terms of investigating the effects of the Coronavirus pandemic on drug demand.