The Comparison About Index Investment Portfolio Optimization Based On Data Mining

We present domestic and foreign optimized index investment literatures, and discover that more useful and effective index tracking information and patterns in mass historical data become new research direction of index investment. However, how to resolve the problem between rich data and poor knowledge? A new breakthrough is data mining to resolve the puzzle. Data mining extracts the process of implicit,unknown and potentially useful information and knowledge, and then deeply handles practical application of data from substantial,incomplete,noisy,ambiguous information. Therefore, this dissertation summarizes 5 kinds of optimized index investment based on data mining. We introduce the theory of data mining methods into index investment research, on the one hand data mining provides innovative solutions for decision-making investment, on the other hand the application for data mining opens up a new field. Consequently, observed from the perspective of quantitative analysis, we can identify the applicability of indexing investment approach by using data mining theories and methods in China's capital market, and can possess the important theoretical significance and academic value by systematic study.Index investment is a kind of passive investment strategies, and the initial standpoint originates from buy-and-hold strategy, it gains the average market return for portfolios by duplicating and tracking benchmark. And index investment has neither the excess risk nor restrictions on timing, as a result, with the deepening development of China's securities market, index tracking turns into the main form of index investment research by more and more theoretical and practical concerns. We can generally classify the index investment methods into three categories:full replication, optimized sampling, and sampling replication. Owing to reasonable and available structure principle of optimized sampling method, the emphasis of this dissertation is how to construct the optimal index investment portfolio by using the optimized sampling.In the empirical design and test of the index portfolio optimization, first, according to the historical data of SSE180 Index and constituent stock from January 5,2009 to June 30,2010, under the limitation of a series of constraints, such as:stock concentration, transaction costs, fund size, market impact cost, the minimum transaction volume, we choose 180 constituent stocks of SSE180 Index as the optimized objects, and achieve optimal holding weights and optimal holding the numbers of 180 constituent stocks by minimizing the objective function optimization (tracking error), then technically construct the optimal index portfolio. Second, we validate and analyze the tracking benchmark, comparing performance and superiority. Finally, we conclude the different consult from various indicators based on data mining methods and also verify the data mining method can not simply compare which one is better.