Research On Intelligence Optimization Algorithms And Its Application In Financial Field

The financial industry is the core of modern economy,and its good operation is a good guarantee for economic construction and social operation.Modern finance is becoming more and more technical,and complex mathematical tools are used in research and practice.As an interdisciplinary subject of finance,mathematics and computer,financial mathematics is originated,applied,and developed rapidly in the new century.The classic problems in the field of financial mathematics are stock forecasting,investment portfolios,and risk management.These problems are effectively solved through financial mathematical models and intelligent algorithm optimization.Bio heuristic computation is a series of intelligent optimization algorithms generated by simulating the evolution mechanism of natural ecosystems.Like natural ecosystems,biological heuristic algorithm can solve complex optimization problems through its own evolution,and has excellent performance in solving NP problems such as TSP,multi-objective,high-dimensional and constrained optimization problems.Since the 1980 s,a large number of scholars began to study biological heuristic algorithms and took the lead in applying them to solve financial optimization problems,including investment portfolio,stock forecasting,risk management and other fields.In the application of financial optimization,the problem of multi-objective optimization is often encountered,and there are many constraints at the same time,so financial optimization scholars are committed to optimize the solution of multiple objectives.Meanwhile,the research of artificial intelligence optimization method based on biological heuristic is in the ascendant,and it is worth further understanding in multi-objective algorithm improvement and algorithm financial optimization application.Therefore,the focus of this thesis is the bio heuristic algorithm and its application in the field of financial industry.The research includes the LSTM model for noise reduction of stock price forecast data,the bio heuristic optimization algorithm based on root growth model,the risk management model optimization based on root algorithm and the multi-objective model optimization of investment portfolio asset allocation based on multi-objective bio heuristic root algorithm.More details and the innovative achievements are listed below:(1)Research on stock price forecast based on multi-optimal combination wavelet transform LSTM modelFirst of all,the process of stock price forecasting is studied,and it is clarified that stock price forecasting is one of the core issues in the financial industry.In the process of prediction,considering the importance of data in model training and the impact on training accuracy,this section discusses the problem of raw data preprocessing.Wavelet transform has a good effect of noise reduction and can solve the high noise problem in financial data.The main innovation of this section is to propose an LSTM model based on multi optimal wavelet transform by improving the threshold selection of wavelet transform.The experimental results show that the multi-optimal wavelet transform model solves the problem of data noise to a certain extent,and the fitting effect of the actual data is good,which is helpful to improve the accuracy of stock price prediction.(2)Research on intelligent algorithm based on root growth modelBased on the study of heuristic algorithm and the characteristics of plant root growth,a plant root growth model was established.On the basis of this model,a single target root algorithm(RA)based on individual behavior is proposed.RA algorithm can automatically and dynamically adjust the direction of root growth by establishing a root evaluation system based on auxin,which is a new type of bioluminescence algorithm.Compared with the classical single target algorithm,RA algorithm is superior to other algorithms in low dimension(d = 20)and high dimension(d = 100),and has good convergence and robustness.Inspired by the group evolution model and the group population idea,based on the single objective root growth algorithm,the multi-objective root growth algorithm(MPMORA,multi population multi objective root algorithm)is proposed by adding multi-objective optimization strategies such as P strategy,GG standard,non-dominated sorting and crowding distance.Through the experiment on the multi-objective standard test function set,it is proved that the MPMORA algorithm can deal with the multi-objective optimization problem with constraints,and has good convergence and uniformity,which is worthy of further study.(3)Research on optimization model and algorithm of investment risk managementThe optimization of RA algorithm in risk management model is realized.First,the purpose of risk management is briefly described,and many characteristics of investment risk management are analyzed in detail.For the purpose of minimizing risks,a distributed decision model considering the level of investment risk is established,and RA algorithm is used for model optimization test.The simulation results show that RA algorithm has better convergence speed and progress in optimizing risk management compared with PSO,GA and ABC algorithm,which further verifies the effectiveness and superiority of RA algorithm.(4)Research on multi-objective portfolio optimization model and algorithmAccording to practical problems,this section proposes a three-objective return-risk-cost model based on the classic portfolio model.With its characteristics,MPMORA algorithm is applied to asset the allocation of portfolio,design the coding,process the optimization,process the constraints,and compare the performance with other classical multi-objective algorithms.Based on the daily historical data of 12 kinds of assets of Shanghai Stock Exchange and the real data collected by the monthly rate of each kind of stock from January 2010 to December 2016,the simulation experiment is carried out.The results show that MPMORA has the ability of solving complex multi-constraint problems,and it has faster convergence speed and higher particle uniformity than the existing classical multi-objective bio heuristic algorithm.Therefore,the algorithm proposed in this thesis provides a better solution for the optimal decision-making of portfolio.