Portfolio Construction Based on Time Series Clustering Method Evidence in the Vietnamese Stock Market

Portfolio optimization is indeed a crucial and highly pertinent topic in finance. During the process of constructing an investment portfolio, investors typically face with two important decisions on portfolio selection and portfolio allocation. This article aims to delve into the realm of portfolio optimization by applying the time series clustering method to historical data of stock returns on the Vietnamese stock market. The VNIndex (VNINDEX) from January 2, 2013, to December 31, 2022 was collected for analysis. About methodology, we utilize hierarchical clustering, which can be applied to a single time series (UTS) or multiple time series (MTS). The resulting portfolio was constructed by selecting stocks from these clusters based on their Sharpe ratio, a measure that finely makes balance between the risk and return. Afterward, we determine the optimal portfolio weights using the Markowitz’s Mean - Variance model, enabling us to create two distinct portfolios: MV_UTS and MV_MTS. Furthermore, our analysis encompasses dynamic evaluation using a rolling window, shedding light on how different methods respond to events impacting financial markets, such as crises. This study provides valuable insights into the application of clustering techniques in portfolio optimization and elucidates the performance and sensitivity of these methods under diverse market conditions.