%0 Journal Article
%T 基于机器学习的选股模型及投资组合研究<br>Stock Prediction Method Based on Machine Learning and Portfolio Research
%A 卢相刚
%A 王泽仁
%J Hans Journal of Data Mining
%P 43-63
%@ 2163-1468
%D 2024
%I Hans Publishing
%R 10.12677/HJDM.2024.141005
%X 股票是金融市场重要的组成部分，其变化存在着一定的内在规律，但是也受到多种因素的制约与影响。因此，如何能够选取好的股票进行操作也成为了很多从业者的研究方向。传统的选股策略有两种，一种为多元回归法，其缺点是对于极端值较为敏感，极端值的存在会影响回归结果，另一种是多因子打分法，其缺点是需要人为给定各个因子的权重，主观性对选股结果有很大影响。本文使用基于决策树的Adaboost模型进行选股，并且构建了投资组合的优化模型，有效提升了投资的收益率。本文的主要工作包括：(1) 建立股票特征指标库，选取更能解释模型的特征指标并对其进行有效性分析和相关性分析；(2) 构建基于决策树的Adaboost选股模型，对模型参数进行优化并且对模型的泛化能力进行评估；(3) 对马科维兹的投资组合模型进行改进，提出一种新的投资组合模型，使得能在降低总体风险的同时将投资收益维持在一个相对高的水平。<br />
Stock is an important part of the financial market. It has certain intrinsic laws of change, but it is also subject to and influenced by many factors. Therefore, how to select good stocks for trading has become a research direction for many practitioners. There are two traditional stock selection strat-egies: one is the multiple regression method, which is sensitive to extreme values, the presence of which will affect the regression results; the other is the multi-factor scoring method, which requires artificially assigning weights to each factor and has a great impact on the stock selection results. This paper uses Adaboost model based on decision tree for stock selection and constructs an opti-mization model for investment portfolio, which effectively improves the investment return. The main work of this paper includes: (1) Establishing a database of stock characteristic indices, select-ing characteristic indices with strong explanatory power, and conducting validity analysis and cor-relation analysis; (2) Constructing an Adaboost stock selection model based on decision trees, opti-mizing model parameters, and then evaluating the effectiveness and generalization ability of the model; (3) Improving Markowitz’s portfolio model and proposing a new investment portfolio model, which can reduce the overall risk and keep the investment return at a relatively high level.
%K 选股模型，机器学习，投资组合<br>Quantitative Stock Selection
%K Machine Learning
%K Portfolio
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=79894