Multi-Algorithm Fusion Pharmaceutical Sales Forecasting Mode

Abstract:
In recent years, China's pharmaceutical e-commerce industry has developed rapidly, and now the industry has entered a critical stage of development. However, when the merchants stationed in the pharmaceutical platform, there are still many problems with the recognition of the pharmaceutical e-commerce and the positioning of the sales target. Many businesses are facing the risk of continuous loss and bankruptcy in the increasingly fierce competition due to their vague positioning and unclear strategic layout. Therefore, this paper mainly analyzes the sales data of merchants staying on the medical platform, and then makes sales forecasts. The sales data involves the number of new users on the day, so first use the time series model ARIMA to predict the number of new users, and then the predicted number of new users is put into the data to predict sales. Time series models ARIMA, LSTM, XGBoost are used as the base models and LightGBM is used as the final prediction model to predict merchant sales using Stacking strategies. This method concentrates the advantages of each model, greatly improves the accuracy of sales forecasts for pharmaceutical merchants, and is of great significance to merchants' decision-making.