Research And Implementation Of Monthly Effect Enhance Mean Variance Model Based On PyCUDA Architecture

With the continuous improvement of GPU parallel computing performance,more and more computation-intensive programs begin to try to use GPU parallel computing capability to improve processing speed.CUDA is a GPU-based parallel computing platform,which can improve the parallel computing speed of large-scale data.At the same time,in the field of financial quantification,it is a good choice to apply CUDA in the field of financial computing,as more and more computations are involved and more and more real-time requirements are required for data calculation and analysis.Considering that the classical mean-variance portfolio model in the field of financial quantification has A low degree of fit when applied in the actual market,this paper improves and strengthens the mean-variance model by using Python and combining the advantages of PyCUDA parallel computing from the perspective of the unique monthly effect of a-share securities market.The specific work is as follows:First,this paper studies the key technologies related to CUDA programming,and then explains the advantages of the PyCUDA module in Python over other cud A-capable programming languages.Then,the quantitative indexes commonly used in data analysis of monthly effect are clarified,and the implementation algorithm of PyCUDA version is given for the indexes suitable for parallel computing.Then,the time-consuming effects of using Python serial computing program and PyCUDA parallel computing program on different data sizes are compared.Then it introduces the theoretical basis of mean variance model and makes derivation and analysis.Finally,to improve the concrete representation of model,and the selection of experimental data to illustrate the use scope,detailing how to make use of monthly effect reinforced concrete implementation steps of variance model,analyzes the reasons of the enhancement effect,further puts forward the improved algorithm of two kinds of enhancement effect is much better,After empirical analysis,the feasibility and effectiveness of the two enhanced algorithms are verified,and the investment suggestions for different types of investors are given.