Time Series Data Forecasting: Alcohol Sales Prediction Based on Prophet

Abstract

Sales forecasting plays an important role in a business’s operation. It points out the outline of future development and customer targets for companies. Due to the fact that sales forecasting is mostly based on large amount of past data, machine learning has a successful implementation in this field. This paper employed Prophet, a package specialized in dealing time series, to forecast the alcohol sales amount across the USA. The data showed significant trends and holiday effects that fit with the algorithm behind Prophet. By adjusting the parameter functions, the model generated by Prophet demonstrated impressive ability on forecasting the third year’s data based on the previous two. A variety of performance indicators, including Mean squared error (MSE), Mean absolute percent error (MAPE), and Median absolute percentage error (MDAP) and so on, were used to evaluate the predicted results The result of this study implies that machine learning has a huge potential in the field of sales forecasting.