The Study Of Portfolio Optimization By Fundamental Network Analysis

Stock market is an important part of modern financial market and one of the most important investment avenues for investors’ asset management.How to choose a suitable stock portfolio for investment in the stock market has been an important issue of concern for stock investors and investment institutions.Portfolio optimization research is to answer the above questions.Generally speaking,the process of portfolio optimization can be decomposed into two:the stock selection process and the asset allocation process of stocks after stock selection.The mean-variance optimization(MVO)model proposed by Markowitz(1952)laid the foundation of modern portfolio theory and provided the answer to the second stage,i.e.,the question of how to allocate equity asset weights under the assumption that the probability distribution of future stock returns is known.The MVO model has become the standard approach to portfolio optimization in the financial industry.However,the model suffers from poor out-of-sample performance,over-concentration of funds,and inability to effectively diversification.For more than 70 years,Considerable effort has been devoted to handling estimation error to address the instability of the Markowitz model,but these advances still do not prevent them.In addition,the MVO model does not involve the stock selection stage,and only uses the statistics of stock prices as the estimates of stock returns and risks,according to the theory of efficient market hypothesis.However,when the stock market only relatively efficient,there is a problem of information deficiency in using only price data.Since the drawbacks of classical portfolio models and the widespread use of complex network methods in financial market research,this work proposes a networkbased portfolio optimization framework called "Fundamental Network Analysis"(FNA)framework.The FNA framework consists of two processes of portfolio optimization at the same time.In the stock selection process,stocks with good expected future returns are selected from the whole market based on fundamental analysis to form a stock pool;in the asset allocation process,the FNA model is applied to allocate stocks in the stock pool based on the principle of diversification.Specifically,first,compared with the MVO model,FNA not only avoids the estimation error and information deficiency problems of the MVO model,but also allocates assets based on the topological nature of the stock network and allocates different weights to stocks with different risks,which can achieve the goal of truly effective diversification.In addition,based on the clustering characteristics of the financial statement structure correlation of listed companies,the use of cluster analysis(similar characteristics within groups but different characteristics between groups)forms a different industry classification criterion from the traditional one.This approach allows better distinguish different financial and operational strategies among listed companies.The main contents and results of this work are as follows.Chapter 1 introduces the research background and research significance of the problem of portfolio optimization,as well as the research content,research framework and innovation points of this paper.Chapter 2 includes a literature review related to this paper.First,this paper fully reviews the relevant literature in terms of the development and shortcomings of classical portfolio models,the application of complex network methods in financial markets,the study of fundamental analysis in portfolios and the significance of entropy theory in portfolios,and summarizes and briefly describes the research methods and innovative points of this paper.Chapter 3 introduces the theoretical foundation and research framework of FNA:The theoretical sources and applications of financial network analysis are described;then the theoretical basis and economic significance of fundamental-based financial networks are introduced,and the empirical analysis of fundamental clustering is used to further demonstrate that the clustering characteristics based on fundamental information can bring about a reduction in the correlation of stock returns between clusters,providing a basis for application to diversified portfolios;in addition,the summarizes the existing asset allocation methods and their drawbacks;finally,specifies the strategy and modeling process of applying the fundamental network analysis model to portfolios,providing a guiding framework and analytical basis for the empirical analysis in the next chapters.Chapters 4 to 6 will construct network models for different regions with different fundamentals based on different fundamental objectives and fundamental variables based on the fundamental network analysis framework for empirical analysis.Chapter 4 introduces the empirical analysis based on the FNA framework based on the application of the aggregate return on equity network model of Chinese listed companies for portfolio optimization:the characteristics of return on equity network stock clustering based on the three factors of DuPont decomposition are analyzed,different from the traditional industry classification criteria are obtained,the network structural entropy is used as a measure of portfolio diversification,and the network clustering for diversified asset allocation is verified The rationality of network clustering for diversified asset allocation is verified.In order to demonstrate the effectiveness and flexibility of the fundamental network analysis framework,Chapter 5 conducts portfolio optimization analysis based on the incremental network model of return on net assets and the incremental network model of operating profit of U.S.listed companies:the differences with the model in Chapter 4 are different fundamental indicators,different data and different risk metrics and asset allocation methods.Chapter 5 uses incremental return on equity and incremental operating profit to construct a fundamental network for analyzing the characteristics of the U.S.stock market,and applies network efficiency as a metric to measure network robustness,resulting in a series of portfolios similar to the efficient frontier,and compares this model with the efficient portfolio of the MVO model.MVO model;the operating profit incremental network model performs due to the MVO model in a crisis market environment;the empirical results prove the validity of both models.Chapter 6 is an analysis of the application of the P/E-based network model to portfolio optimization.Chapters 4 and 5 mainly examine the impact of listed companies’ profitability on the portfolio.Chapter 6,on the other hand,considers the impact of firm value and firm size on the portfolio.Through the P/E distribution,complex networks are built for large,medium and small capitalization stocks respectively,and the local systematic risk of the portfolio is measured by the entropy of the network structure.In the empirical analysis of the U.S.S&P 1500 constituents,we find that the method optimizes best for mid-cap stocks and achieves a higher investment return than the traditional portfolio optimization solution.Chapter 7 is a summary of the work,which summarizes the main results and policy recommendations of this paper,and also sorts out the future research directions.