摘要
摘要
世界经济空前发展，资本市场也随之不断发展和完善，证券投资已经成为人
们生活的一部分，股票市场作为经济金融的晴雨表，备受投资者青睐。量化投资
作为近些年兴起的投资方式，随着计算机技术的提升和理论框架的完善逐渐进入
投资者的视野。此外，量化私募基金的数量和规模也在近几年空前发展，量化产
品不断创新，收益颇丰，尤其是以中证 500指数为基准的指数增强类产品，表明
投资者更加偏好投资成长类中小盘股。在此背景下，对量化投资策略进行深入研
究，对市场与投资者具有重要意义。
为了丰富量化投资策略内容，验证策略的有效性，本文以中证 500指数成分
股为研究对象构建量化投资策略。策略主要有两个主体部分，分别使用多因子打
分选股模型和长短期记忆（LSTM）神经网络预测模型对股票池中的股票进行筛
选，最终对输出结果进行回测和评价。具体来说，首先通过因子有效性分析和相
关性分析对股票候选因子池进行初步筛选，并确定基本面有效因子和技术面有效
因子；其次，再运用主成分分析对有效因子进行降维处理，形成最终因子池；接
下来，基于最终基本面因子，每月进行多因子打分选股，筛选出基本面表现优异
的股票；然后，基于最终技术面因子针对上一环节筛选出的优秀股票，应用 LSTM
模型对其进行周度预测；最后，对基本面表现优秀且预测未来会上涨的股票进行
交易，每日统计持仓市值并形成交易记录，根据交易回测结果，计算相关评价指
标并据此对该策略进行评价。
基于上述研究内容，本文利用 2011年至 2018年的数据进行有效因子检验和
筛选，2019年至 2020年数据进行交易回测，验证策略效果。回测结果表明，本
文所构建的策略能够稳定获取超额收益，能够在市场下跌时起到缓冲作用，在市
场上涨时增厚收益。近两年中证 500指数收益表现已然非常优秀，在此基础上进
?
行多因子打分选股确实获取了比较可观的收益，再选股的基础上添加股价预测
之后，策略收益有一定的上涨。综合来讲，各个评价指标都比较向好，充分说明
本文所构建的量化投资策略是一个比较好的策略，具有一定的投资参考价值，同
时也侧面验证了市场的不完全有效性。
本文将基本面分析和技术分析内在结合，在考虑企业价值的同时，捕捉市场
I

摘要
短期交易信号，缓冲下跌风险，增厚策略组合收益。在市场不完全有效的前提下，
通过计算机的快速运算和大数据分析，能够有效捕捉市场交易信号，这正是量化
投资备受青睐且快速发展的原因之一，本文基于量化投资理论，结合深度学习算
法构建量化投资策略，丰富量化投资策略研究内容，在量化投资发展快速发展的
背景下，为投资者提供投资思路参考。
关键词：量化投资策略；多因子打分选股；LSTM模型；主成分分析
II

Abstract
Abstract
With the unprecedented development of the world economy, the capital market has
also been developed and improved, and securities investment has become a part of
people's life, and the stock market, as a barometer of economy and finance, is highly
favored by investors. Quantitative investment, as an investment method that has
emerged in recent years, has gradually entered investors' vision with the improvement
of computer technology and theoretical framework. In addition, the number and scale
of quantitative private equity funds have also developed unprecedentedly in recent
years, and quantitative products have been innovated and yielded good returns,
especially index-enhanced products with CSI 500 index as the benchmark, indicating
that investors prefer to invest in growth small and mid-cap stocks. In this context, an
in-depth study of quantitative investment strategies is of great significance to the market
and investors.
In order to enrich the content of quantitative investment strategy and verify the
effectiveness of the strategy, this paper constructs a quantitative investment strategy
based on the component stocks of China Securities 500 index. There are two main parts
of the strategy: a multi-factor scoring stock selection model and a long and short-term
memory (LSTM) neural network prediction model are used to screen the stocks in the
stock pool, and the output results are finally back-tested and evaluated. Specifically,
firstly, the stock candidate pool is initially screened by factor validity analysis and
correlation analysis, and the fundamentally valid factors and technically valid factors
are identified; secondly, the valid factors are then downscaled using principal
component analysis to form the final factor pool; next, based on the final fundamental
factors, monthly multi-factor scoring stock selection is performed to screen stocks with
excellent fundamental performance; then Then, based on the final technical factors, the
LSTM model is applied to the stocks screened in the previous section to make weekly
forecasts; finally, the stocks with excellent fundamental performance and predicted to
rise in the future are traded, and the market value of the positions are counted daily and
III
。。。以下略