Abstract
This paper improved the forecasting performance of Fuzzy Time Series (FTS) by incorporating Cat Swarm Optimization Clustering (CSO-C) and Particle Swarm Optimization (PSO) with FTS. In the three stages of FTS, clustering technique (CSO-C) was applied at the first stage (fuzzification) to objectively partition universe of discourse and eliminate the need for defining universe of discourse. At the second stage (determination of fuzzy relation), disambiguated fuzzy relations were obtained using Fuzzy Set Group (FSG). PSO was implemented at the last stage (defuzzification) to optimize by tuning weights assigned to fuzzy sets in a rule. Auto Regressive Integrated Moving Average (ARIMA) and Seasonal Auto Regressive Integrated Moving Average (SARIMA) are the most common methods used locally to forecast accident data. However, they are not applicable to every season even though they also yield a good result. Hybrid fuzzy time series method is applicable regardless of season. Minna-Bida highway data forecast gave a RMSE of 2.494 and MAPE of 0.386% while Yusuf et al (2015) gave RMSE of 2.894 and MAPE of 0.456%. It implies that; this method outperformed other conventional and non-conventional methods existing in literature using Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) as performance metrics.

National Library of Nigeria
Association of Nigerian Authors
Nigerian Library Association
EagleScan
Crossref