Research And Implementation Of Data Analysis Technology In Quantitative Transactions

Due to the rise of quantitative investment,the analysis of stock price time series has become increasingly important.In financial quantification,the development trend of future stocks can be effectively predicted by searching for similar stock price time series and matching with financial models.In order to solve the problem of existing search and model matching algorithms,which are high time complexity and low accuracy.This paper propose a similar stock identification method based on dynamic regular distance,a hybrid financial model matching algorithm based on key point extraction to improve the efficiency and accuracy of similar sequence searching and financial model matching problem.We design and implement a set of data acquisition and analysis platform for quantitative transactions.The main research contents of the thesis are as follows:(1)A similar stock series sear-ching method based on DTW.Firstly,our method extracts the trend characteristics of the query sequence based on the results of APCA,extreme point and other dimension reduction methods.Second,we use the feature as a key to search for sequences with the same characteristics to form a candidate set.Then,for each sequence in candidate set,we filter the important points and add them to the list of feature points to better preserve the shape of the original sequence.Finally,DTW algorithm is used to find similar time series.Experimental results show that our method can improve search efficiency and accuracy.(2)A hybrid financial model matching algorithm.Firstly,we extract the financial model feature points in the stock time series according to the financial model's value and time requirements based on the dimensionality reduction results to reducing the influence of noise and small fluctuations that do not affect the trend.Then we fix the original financial model and filter the time series through the model so that we can find similar time series that satisfy the financial model quickly and accurately.Experiments with real-world financial data sets show that this method has higher accuracy and less time than traditional methods.(3)Data analysis platforms for quantitative transactions.The system is divided into three modules:web server,calculation engine,data extraction engine and data clean module.They are responsible for accepting user requests and execute related business logic,finding similar time series and matching financial models,maintaining database and providing read and write interfaces and checking the data quality while improving data availability.