A Study on Fuzzy Time Series Forecasting using Machine Learning Techniques

Abstract

Over the past few decades, many researchers have been putting their effort to develop non-deterministic fuzzy time series (FTS) models using the traditional fuzzy set. However, contrasting to the traditional fuzzy set, the hesitant fuzzy set allows a set of membership values to each element. Thereby it glorifies the chances to capture the fuzziness and uncertainty due to randomness better than the traditional fuzzy set. Motivated by this, the present study proposes a novel hesitant FTSF model (HFTSF) using support vector machine. Despite of over last two decades of research on fuzzy set theory, researchers have only concentrated on modeling non-statistical uncertainties. Normally non-statistical uncertainties are modeled using fuzzy set theory where the partial truth of an event is denoted by assigning the degree of belongingness of an element. However, the fuzzy set is not capable to efficiently model the hidden chances of occurrence of this event. At the same time, the use of statistical uncertainties can be a better choice to capture the chances of occurrence with a hidden probability of an event in future. As per the above concept, the accuracy of the forecasting model can be more satisfactory if the probability and fuzziness are expressed at the same time to represent both the statistical and non-statistical uncertainties. Motivated by this the present research proposes a novel probabilistic intuitionistic fuzzy time series forecasting (PIFTSF) model using support vector machine (SVM) to address both uncertainty and non-determinism associated with real world time series data. newline