近年來，在股市投資的方法上已經有許多研究人員提出各種投資組合策略，考量到提高投 資組合的獲利，研究人員對於股票之間的連動性非常感興趣。為此，本篇論文透過時間相 似性度量指標之 t-隨機鄰域嵌入法 (DTW based on t-SNE)及階層分群法，將多維度的股 票市場資訊做為特徵，對資料做視覺化及分群處理，並運用熱圖呈現每檔股票在不同時間 段上的狀態變化。本研究之成果是希望透過本研究所提出之“股票狀態視覺化熱圖”，視覺 化不同股票間可能存在之相關性，並找出具有相似狀態變化之股票間的穩定或新興集群。 並透過股票的狀態變化進一步分析資金的流動變化。 ;In recent years, numerous researchers have proposed various investment portfolio strategies in the field of stock market investing. To enhance portfolio returns, researchers have shown a keen interest in understanding the interrelationships between stocks. In this paper, we uti lize the DTW (Dynamic Time Warping) based on t-SNE (t-Distributed Stochastic Neighbor Embedding) algorithm, along with hierarchical clustering, to visualize and cluster multidi mensional stock market information. We extract multiple dimensions of data as features and present the state changes of each stock across different time periods using a heat map. The main objective of this research is to develop the “Stock State Visualization Heat Map”, which visualizes the potential correlations among different stocks and identifies stable or emerging clusters of stocks exhibiting similar state changes. Furthermore, by analyzing the changes in stock states, we aim to gain insights into the flow of funds within the market.
