Fuzzy time series forecasting with picture fuzzy clustering (FC-PFS) and picture composite cardinality (PCC)

Abstract

The concept of fuzzy time series forecasting is used to make a decision. Fuzzy time series classifies data into several subintervals. In the fuzzy time series forecasting method proposed by Qiang Song and Brand S. Chissom, the fuzzy time series data clustered into some static long intervals. In this paper, we modify the method using Picture Fuzzy Clustering (FC-PFS) to get the optimal clusters. To get the suitable number of clusters with the best quality, Picture Composite Cardinality (PCC) is used. The modified method is applied to predict Indonesia Composite Index (ICI) in May 2021. The AFER value from this application is 0.55% which means that the modified method is categorized as very accurate for predicting. In other words fuzzy time series forecasting with FC-PFS has good performance. We also show that in this study case fuzzy time series forecasting with FC-PFS is better performance than fuzzy time series forecasting with Fuzzy C-Means Clustering (FCM).