HYBRID TIME SERIES MODEL FOR PROCUREMENT FORECASTING IN ENTERPRISE RESOURCE AND PLANNING (ERP) SYSTEM: A CASE STUDY

Abstract

Machine learning in an Enterprise Resource and Planning (ERP) system allows a business to save, organize, and analyze data to make better decisions and obtain valuable views that were previously inconceivable. Forecasting is one of the most common applications of machine learning in ERP. This is a crucial part of a company’s financial planning, which looks to the future based on historical data. This paper presents a novel hybrid model to procurement forecasting based on time series data found in ERP system. The proposed model is a hybrid model of ARIMA (autoregressive integrated moving average) and LSTM (Long-Short-Term Memory). The proposed model combines the linearity and non-linearity components for better predictions. In the first step, the linear components are represented using an ARIMA model. In the next step, the ARIMA model residuals are utilized as input to the LSTM model. By modeling the residual series, the LSTM model is used to train the nonlinear tendency. The evaluation of the proposed model is measured using the Mean Absolute Error (MAE) and Root Mean Squared Error (RSME).