Clustering Financial Time Series Based on Perceptually Important Point Calculation

• Main Authors: Tsu-An Chao, 趙子安
• Other Authors: 羅淑娟
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/6p5x8z

碩士 === 國立臺北科技大學 === 資訊與運籌管理研究所 === 101 === It is usual to observe time series data appearing in many fields such as science, engineering, business, finance, economic and health care. Many researchers use data mining skills to cluster time series data in their studies. Time series data easily become huge and highly dimensional with time cumulated. It is not easy to analyze time series data directly, due to the complexity of time series data. Therefore, in this study we attempted to reduce the dimension of time series before clustering the data. Our study used the Perceptually Important Point (PIP), Dynamic Time Warping (DTW), and Hierarchical Clustering to construct a systematic procedures to search the stocks with same trend for investors. The results showed that PIPs have better similarity and clarity than those without noise reduction. Investors can get more detail levels of the clustering results from hierarchical clustering chart if needed. Based on our procedure, investors can clearly and easily observe those stocks with the similar trends under different clustering levels.