Localized trend model for stock market sectoral indexes movement profiling

Abstract

Previous studies have found that one of the main challenges in the area of time-series analysis is the lack of ability to reveal the hidden profiles of observed dynamic systems. Therefore, this study applies an adaptive clustering method named the Localized Trend Model to extract and group dynamic recurring trends from trajectories of multiple time-series data to expose their underlying profiles of movement. Consequently, in this research localized dynamic profiles of movement between sectoral indexes from the Indonesia stock exchange market in the year of 2016 are extracted, analyzed and utilized to predict their future values as a case study. Results of conducted experiments confirmed that the employed method is capable to perform movement profiling for the Indonesia sectoral indexes and be of help to better understand their imperative basic behavior. Furthermore, the study has also verified the proposition that the ability to better understand profiles of movement in a collection of time-series data would benefit to increase prediction accuracy.