论文题目：基于遗传规划算法的量化交易策略设计——以沪金期货为例
学科专业：金融硕士
学位申请人：李皓燃
指导教师：朱敏
摘要
在我国金融市场中，传统的基本面分析理论和技术分析理论深受投资者的青
睐，近年以来随着量化投资的兴起，量化分析理论也越发受到投资者追捧。尽管
不同理论在各自的适用领域中都能产生很好效果，但三种理论之间始终缺乏有效
的结合点，不能将各理论的优势同时发挥出来。近年来，在宏观经济周期、国家
政策和投资者情绪的多重影响之下，我国的基金规模开启了大幅上涨的趋势，在
基金总量上涨的同时市场中投资策略呈现高度的一致化，具体表现为股市机构的
抱团现象和期货CTA策略同质化现象等。在金融市场中，大量资金采取同质化策
略进行投资的结果会导致拥挤交易现象，资金赚取收益的能力也会大幅减弱。
基于此，本文通过梳理大量相关文献并基于实际的投资经验，重构遗传规划
算法机器学习模型，创造了一种全新的启发式策略投资体系。这种新的投资体系
框架之下，本文实现了从基本面分析到机器学习量化投资的有机结合，并且提供
一种能一定程度上减少市场上的策略拥堵风险的新思维。
具体实证中，本文以沪金期货为实验研究标的，选用黄金基本面分析筛选后
的指标，对2008年沪金期货上市以来至2021年的数据进行分析，并对模型结果
分样本内和样本外两个区间进行回测。实验实现了基于沪金期货的基本面分析与
技术分析量化投资策略，启发式随机产生的策略因子能有效减少策略拥堵风险。
在实验的最优模型样本外的回测中，量化投资策略达到了年化收益 23.80%，夏
普比率2.686，最大回撤-7.26%，显著优于同期市场表现，验证了模型的有效性。
因此，第一，本文的研究成果可以为投资者在如何有效结合基本面分析、技
术分析和量化投资方面提供一定的参考；第二，本文的研究成果在帮助投资者规
避策略拥堵方面具有一定的参考意义。
关键词：基本面量化；遗传规划；量化策略；随机搜索
I

TITLE: Quantitative trading strategy design based on genetic programming algorithm----
Take Shanghai gold futures as an example
MAJOR: master of finance
APPLICANT:Haoran Li
SUPERVISOR:Min Zhu
Absctract
In my country's financial market, traditional fundamental analysis theories and
technical analysis theories have been favored by investors. In recent years, with the rise
of quantitative investment, quantitative analysis theories have become more and more
sought after by investors. Although different theories can produce good results in their
respective fields of application, the three theories lack effective integration points, and
the advantages of each theory cannot be brought into play at the same time. In recent
years, due to the multiple influences of economic cycles, macro policies and investor
psychology, the size of different funds in my country has risen sharply. While the total
amount of funds has risen, investment strategies in the market have shown a high degree
of uniformity. This is the grouping phenomenon of stock market institutions and the
homogenization of futures CTA strategies. In the financial market, the result of a large
amount of funds adopting a homogenization strategy for investment will lead to
crowded transactions, and the ability of funds to earn income will also be greatly
reduced.
Based on this, this paper combed through a large number of relevant literature and
based on actual investment experience, reconstructed the genetic programming
algorithm machine learning model, and created a new heuristic strategy investment
system. Under the framework of this new investment system, this article realizes an
organic combination of fundamental analysis to machine learning quantitative
investment, and provides a new thinking that can reduce the risk of strategic congestion
in the market to a certain extent.
In specific empirical research, this paper uses Shanghai gold futures as the
experimental research subject, selects indicators selected from gold fundamental
analysis, and analyzes the data from the listing of Shanghai gold futures in 2008 to 2021,
and divides the results of the model into in-sample and out-of-sample. Back-testing in
II

two intervals. The experiment realized the quantitative investment strategy based on
the fundamental analysis and technical analysis of Shanghai gold futures, and the
strategy factors generated randomly by heuristics can effectively reduce the risk of
strategy congestion. In the out-of-sample backtest of the optimal model of the
experiment, the quantitative investment strategy achieved an annualized return of
23.80%, a Sharpe ratio of 2.686, and a maximum drawdown of -7.26%, which was
significantly better than the market performance over the same period, which verified
the effectiveness of the model.
Therefore, first, the research results of this article can provide investors with a
certain reference on how to effectively combine fundamental analysis, technical
analysis and quantitative investment; second, the research results of this article have a
certain degree of helping investors avoid strategic congestion. D.
Keywords: fundamental quantification; genetic programming; quantification strategy;
random search
III
。。。以下略