Predicting FMCG Indexes using ARIMA, SVM and ARIMA-SVM Hybrid Forecasting Models

Abstract

The fast-moving consumer goods (FMCG) market, a significant player in the global economy, operates in a highly competitive environment with low-profit margins, necessitating an efficient supply chain. The importance of accurate forecasting in this context cannot be overstated. In the Indian market, even a slight improvement in prediction accuracy can profoundly impact the overall economy and ensure food security for the entire nation.

Real-time data sets from December 1993 to November 2023 are used to predict the complex pattern of FMCG market price. In this time series prediction, ARIMA, SVM, and a Hybrid ARIMA-SVM machine learning model are used. These three models are more efficient in forecasting the time series than other methods in statistics. The ARIMA model is efficient in handling linear trends and seasonality in data. The SVM model is efficient in handling non-linear data relationships. The hybrid ARIMA-SVM model combines both strengths and theoretically shows a better outcome.

Our study revealed that the ARIMA machine learning model outperforms the SVM model, and the Hybrid ARIMA-SVM model demonstrates superior accuracy in forecasting. These findings have direct and significant applications for stakeholders in the FMCG industry, offering them a powerful tool for future decision-making.