Sales Forecasting Using Deep Neural Network And SHAP techniques

In today's business world's consumer-centric environment, enterprises seeking good sales performance require a balance between meeting customer demand and controlling the cost of inventory. Good sales forecasts play a crucial role in assisting enterprises to improve. With the development of Artificial Intelligence, increasing methods have been adopted to solve the forecasting problem. In this paper, we propose the Neural Network sale prediction model for predicting Walmart's sales. Moreover, we evaluate our NN model on the datasets provided by the Kaggle platform. Experiments have shown that compared with other machine learning model, our NN model achieves superior performance. Our RMSE metric is 2.92 and 2.58 lower than the Linear Regression algorithm and the SVM algorithm. Furthermore, to mine attributes of different dimensions to make prediction well, we utilize SHAP to interpret our NN model.