Income Forecasting for Multi-Category Product using ARIMA Approach

The rapid development of e-commerce platforms has brought significant changes in consumer behavior and business marketing strategies. In facing increasingly tight competition, some online stores need to predict revenue for the products they have. However, most online stores have more than one category, so it is necessary to evaluate predictions for each category. This problem can be handled by implementing Autoregressive Integrated Moving Average (ARIMA). This approach can be used to handle non-stationarity by having "integrated" components become stationary through differencing. The average MAPE value of all categories of cosmetic products is 12.144%. The development of this prediction system is in accordance with the stages in CRISP-DM. This approach provides a structured framework for carrying out data mining projects with six clear phases.