A novel hybrid time series forecasting model based on neutrosophic-PSO approach

Abstract

This article proposed a new time series forecasting model based on neutrosophic set (NS) theory and particle swarm optimization (PSO) algorithm. The proposed model initiated with the representation of time series dataset into NS using three different memberships of NS, i.e., truth-membership, indeterminacy-membership and falsity-membership. This NS representation of time series was referred to as neutrosophic time series (NTS). It was observed that the forecasting accuracy of the proposed model was highly relied on the optimal selection of the universe of discourse of time series dataset. In this study, this problem was resolved by using the PSO algorithm. The proposed model was verified and validated with three different datasets that included the university enrollments dataset of Alabama, TAIFEX index and TSEC weighted index. Experimental results showed that the proposed model outperformed existing benchmark models with average forecasting error rates of 0.80%, 0.015% and 0.90% for the university enrollments, TAIFEX and TSEC, respectively.