The Research And Implementation Of Graphical Signal Matching In Quantitative Trading

With the rapid development of quantitative transactions, how to use the computer to identify the graphical signal in the K-line graph has become an important issue. At present, computer recognition graphics signals are mostly based on neural networks and fuzzy logic systems. The construction and training process of the neural networks is cumbersome,the neural network which has been trained can only identify a single graphical signal. It is too cumbersome in the practical application. The fuzzy system for the definition of the signal is complex, because you need to define the graphical signal of each K line and the relative position of the adjacent K line. It does not apply to a long time span graphic signal.In this paper, a pattern matching algorithm based on template method is proposed, and the definition and matching methods of graphics signals are studied and experimented. The main work of the thesis is as follows:(1) The classical signal is sorted and classified according to the shape of the graphic signal. The graphic signal is divided into three types,simple graphic signal, oscillating graphic signal and complex graphic signal, and three types of graphic signals are defined respectively. The graphic signal is represented in a 10 * 10 matrix. The trend of the graphic signal is represented by the elements whose weights are nonzero in the matrix. The smaller the weights of the elements, the bigger the distance is.(2) The historical price window is divided by the time window,and the time window is moved in the increasing direction of the transaction time. The price trend in the time window is transformed into matrix representation, and the price trend is represented by the non-zero elements in the matrix. Finally, the matrix is "windowed" and"normalized" to get the historical price window matrix.(3) Calculate the similarity of the graphical signal template matrix and the historical price window matrix, which is the matching result of the graphic signal in the historical price trend. If the match value exceeds the threshold, the match is considered successful. Otherwise, the match fails. The matching result threshold needs to be determined by the experimental results.(4) Based on the above matching algorithm, finish the design and implementation of the graphics matching system. The graphics matching system is divided into three layers: presentation layer, business logic layer and data storage layer. The presentation layer is responsible for showing the matching results and posterior trend statistics results. The business logic layer is responsible for the definition of graphics signals and graphics signal matching implementation, which is the core of the graphics matching system. The data storage layer is responsible for the storage of historical price data and graphical signal definitions.(5) Matching the historical price data with the graphic matching system, and verifying the validity of the matching algorithm. And the posterior trend of the identified graphical signal is statistically counted to verify the validity of the prediction significance of the graphic signal.