A method for fuzzy time series forecasting based on interval index number and membership value using fuzzy c-means clustering

Fuzzy time series forecasting methods are very popular among researchers for predicting future values as they are not based on the strict assumptions of traditional forecasting methods. Non-stochastic methods of fuzzy time series forecasting are preferred by the researchers over the years because these methods are capable to deal with real life uncertainties and provide significant forecast. There are generally, four factors that determine the performance of the forecasting method (1) number of intervals (NOIs) and length of intervals to partition universe of discourse (UOD), (2) fuzzification rules or feature representation of crisp time series, (3) method of establishing fuzzy logic rule (FLRs), (4) defuzzification rule to get crisp forecasted value. Considering, first two factors to improve the forecasting accuracy, we proposed a modified non-stochastic method of fuzzy time series forecasting in which interval index number and membership value are used as input features to predict future value. We suggested a rounding-off range and large step-size method to find the optimal NOIs and used fuzzy c-means clustering process to divide UOD into intervals of unequal length. We implement two techniques (1) regression by support vector machine and (2) neural network by multilayer perceptron to establish FLRs. To test our proposed method by both techniques we conduct a simulated study on eight widely used real time series and compare the performance with some recently developed models. Two performance measures RSME and SMAPE are used for performance analysis and observed better forecasting accuracy by the proposed model.