Application of Automatic Clustering and Fuzzy Time Series Type-2 in Indonesia Composite Index

Abstract : Forecasting the Indonesia Composite Index (ICI) is one of the important efforts in making investment decisions in the capital market. This index helps investors see changes in stock prices directly, making it easier to know whether stock prices are rising or falling. In this research, a combined approach between Fuzzy Time Series (FTS) Type-2 and an automatic clustering algorithm is applied to improve the accuracy of ICI forecasting. The advantage of FTS Type-2 is that it can provide more accurate results than FTS Type-1 because it can express more information. At the same time, automatic clustering is also used because it can partition the universe of discourse efficiently. The data used is monthly data on the closing, highest, and lowest prices of the Indonesia Composite Index from January 2019 to July 2024. The results of this study show an AFER value of 0.6827 and a forecasting accuracy of 99.317%. This shows that the forecasting results using FTS Type-2 and automatic clustering algorithms provide excellent forecasting results. This research has the novelty of using automatic clustering in determining the interval of Fuzzy Time Series Type-2 for ICI forecasting.