Studies On Portfolio’s Optimization Of Chinese Credit Abs Products Based On AI Quantitative Analysis

Asset Backed Securitization(ABS)is an important financing channel for financiers in modern financial markets and a new tool for financial risk management.With the increasing attention paid to ABS in recent years,it provide more opportunities for growth its market scale and its product types.It plays an increasingly important role in promoting the optimization and upgrading of economic structure and serving the real economy.ABS in China market has entered a critical period of construction and development,and product innovation has become an inevitable trend.Along with the gradual maturity of China’s securities market,the new generation of information technology represented by Big Data,Cloud Computing,Artificial Intelligence(AI)has also developed rapidly.Quantitative investment method based on AI can track and monitor the whole market with the help of powerful computing capability,and search for potential trading opportunities in the market in time with accurate evaluation,which can effectively reduce the adverse impact of personal factors on transaction decision-making.Thus,it is timely to integrate AI with financial sector.Under this background,this dissertation introduces AI quantitative investment into ABS,and studies the optimal portfolio selection scheme.Firstly,the first and second chapters of this dissertation elaborate the relationship between ABS and AI,analyze the differences between ABS between China and the United States,and discuss the current application of AI in domestic ABS product portfolio.Secondly,this dissertation investigates the transaction rules and investment costs of ABS in China,with five sets of real transaction scenarios and assumptions from different investors’ perspectives,constructs an ideal yield model.Secondary Transaction;data from 2015 to 2018 in three main market including Shanghai Stock Exchange,Interbank Stock Exchange and Shenzhen Stock Exchange are used to obtain the ideal return under different parameters.Based on L2 norm,the average level and volatility of Sharp ratio in each portfolio of investment strategy are considered.Lastly,Lasso regression,ridge regression,XGBoost,LightGBM,neural network and SVM(Support Vector Machine)are established based on the idea of optimizing the ideal model,and non-zero combination(excluding no investment decision)in AI investment strategy is used.Based on the Sharp Ratio Mean and Sharp Ratio Volatility of Non-zero Portfolio(excluding portfolios without investment decisions)in AI Investment Strategy,the Dis Index is used to quantitative analysis the disparity of the AI performance and the Ideal Return Rate as the Benchmark Performance.The experimental results show that ridge regression model,XGBoost model and Lasso regression have the best performance with the ideal return rate when the structural rate parameters change,and the other parameters are constant.However,when the other parameters change,and the optimal model is Lasso regression,which is less sensitive to the two parameters of repurchase pledge ratio and leverage ratio.The investor who under these two parameters can refer to the results.It further illustrates that using AI method can bring better investment performance to investors.This dissertation mainly uses the mainstream AI models,such as Lasso regression,ridge regression,XGBoost,LightGBM,neural network and SVM,and gives the mathematical or algorithmic basis of each model.Through the construction of ABS product allocation strategy and the empirical study of various models,it is proved that the constructed Ridge Regression has certain advantages in comprehensive.The Fitness index and subsequent comparisons show that Ridge Regression has better stability,and has both theoretical rigor in mathematics and superiority in investment performance.Regarding the change of structural rate,ridge regression still shows an eye-catching performance,which further reflects that when using AI method to analyze investors,we should adopt multi-perspective and multi-model analysis,so as to bring better investment performance.An intelligent optimization model of portfolio based on genetic algorithm is proposed.At present,the research results about asset securitization and portfolio optimization are relatively limited,and most of them are scattered in the asset management of asset securitization.The research on the portfolio optimization of asset securitization mainly focuses on the mean variance portfolio,the resampling portfolio and the portfolio optimization based on the characteristics of the company.The global optimization ability of these optimization methods is weak and the universality is low,so the intelligent optimization model of portfolio based on genetic algorithm is constructed to optimize the capital portfolio globally.