Investigation of a Suitable Hybrid Time Series Model for Predicting Clove Price

Abstract

The prediction of clove prices in domestic markets is affected by non-linear factors, including the monopoly market operational environment. Most of the hybrid time series models used in the prediction of crop prices do not consider the monopoly market operational environment as a nonlinear factor. This study investigated a suitable hybrid time series model for predicting clove price under the monopoly market in Zanzibar, Tanzania. The study conducted desk reviews on existing hybrid time series models and realize that ARIMA-ANN, ARIMA-SVM, ARIMAX-ANN, and SARIMA-NARNN are the most common and efficient models that have been employed to predict crop prices. Based on identified models, the study performed several experiments to investigate the accuracy of these models on predicting clove prices with the monopoly market operational environment as a nonlinear factor. The mean absolute percentage error (MAPE) was used as a performance metric. The monthly average prices of cloves from January 2007 to December 2019 were used to utilise these experiments. Results show that the ARIMA-SVM (MAPE = 0.45%) outperformed the ARIMA-ANN model (MAPE = 0.48%) in predicting clove prices under a monopoly operational market. The study recommended future research to investigate hybrid models for predicting production and planted areas of cloves.