Sales Forecasting Based on AutoRegressive Integrated Moving Average and Recurrent Neural Network Hybrid Model

Abstract

Sales forecasting is essential to a survival of a business in this competitive world. Nowadays with the high volatility in sales revenue, accurate sales predication is a challenge. Hence, recurrent neural networks in forecasting has become popular over the recent years due its data driven nature, flexibility and usage of multiple inputs to identify sequential connection between those. While recurrent neural network individually makes accurate predictions better than traditional statistical forecasting models like autoregressive integrated moving average model, a combination of statistical autoregressive integrated moving average model and recurrent neural networks model enhances performance and it is the most suitable for this domain. Therefore, the proposed hybrid model is capable to identify features of past sales data, statistically predicted data, counts of positive and negative reviews of a particular product for a given period and correlate all of them to yield an accurate sales prediction.