Research And Implementation Of Quantitative Configuration Method Based On Mass Stock Trading Data

With the continuous development of China’s society and economy and the continuous improvement of the financial market system,the financial industry also puts forward new demands for the development of financial informatization.Especially with the rapid development of information technology,the demand of using information technology for data mining,stock screening,quantitative configuration,model back testing,timing and warehouse adjustment in the process of investment analysis and decision-making is increasing day by day.The complex and changeable securities market has produced a large amount of stock trading data.How to make better use of information technology to mine potential information and hidden value to build a portfolio and serve investors in the way of quantitative allocation of portfolio is a field worthy of in-depth research.The research content of this topic mainly focuses on the following three points: on the one hand,it studies how to collect and verify data stably and efficiently in the face of massive stock data;On the other hand,taking the stock calendar effect model as the starting point,this paper studies the clustering algorithm of data mining,selects highquality stocks and constructs high-quality stock portfolio;Finally,the effects of several intelligent optimization algorithms on optimizing high-quality stock portfolio are studied and compared.1.Data collection stage: firstly,this paper investigates and studies the feasibility of massive stock transaction data collection,and a multi-source heterogeneous stock data crawler method based on distributed agent pool middleware is proposed to ensure the stability and accuracy of stock data collection.2.Preprocessing stage: Different from the common clustering algorithms based on stock return or correlation matrix,this paper proposes to cluster stocks by using k-medoids algorithm,DBSCAN algorithm and options algorithm from the characteristic calendar effect of stocks,eliminate inferior stocks and construct high-quality stock portfolio;At the same time,a stock similarity coefficient is proposed as the evaluation standard of k-medoids algorithm.3.Optimization stage: for the allocation of high-quality stock portfolio screened by clustering,this paper takes Markowitz mean variance model and black Litterman model as the theoretical basis,and uses GA,DE,PSO and SQP to optimize the quantitative allocation of portfolio.