股票个性化推荐方法研究

发布时间：2018-01-02 13:06

本文关键词：股票个性化推荐方法研究　出处：《哈尔滨工业大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着我国股市规模的不断发展壮大，股市参与者出现了爆炸式增长的势头。从我国现阶段股民的组成结构来看，大多数投资者是属于专业知识薄弱的中小股民。为了正确引导股市投资者的价值取向，减少盲目投资造成的资源浪费，这就需要为广大中小股民提供切实可行的投资指导。当前的股票推荐方法主要集中在两类，分别为基于股评的在线股票推荐方法和基于数理分析的股价预测模型，它们拥有各自的优势，但缺陷也相当明显，前种方法不能满足股民个性化股票推荐的需求，后种方法应用过程较为复杂，很难被投资者理解并掌握。因此，股票个性化推荐方法的研究成为当下重要的课题。近年来，电子商务呈现蓬勃发展的新局面，，个性化推荐方法被广泛应用于电子商务领域来针对目标用户进行商品推荐。股票可以被看作是一类特殊的商品，基于此，借鉴商品个性化推荐方法的核心思想来构建股票个性化推荐模型是一个可行的思路。本文在充分研究国内外关于推荐方法文献基础上，提炼出了商品个性化推荐方法的核心思想。通过对原有推荐方法进行改良和创新，构建了股票个性化推荐模型。建模思路分为两个步骤：首先，构建股民特征指标体系并运用模糊聚类方法来进行股民群体细分；其次，运用图论和信息最大化保留思想来改进原有推荐方法，建立了基于模糊聚类的股票个性化推荐方法。完成建模过程后，本文首先运用仿真模拟方法来建立运用于实证过程的股民-股票评分数据库，然后，运用编程方法来实现股票个性化推荐流程。最后，将本文推荐方法与基于协同过滤技术的推荐方法和随机推荐方法进行推荐精确度和推荐误差方面的比较分析，得出本文的股票个性化推荐方法具有较高的推荐精度，是一种高性能的实时在线股票推荐方法。
[Abstract]:With the continuous development of the scale of the stock market in China, the participants in the stock market have explosive growth momentum. Most investors are small and medium-sized investors with weak professional knowledge. In order to guide the value orientation of stock market investors, reduce the waste of resources caused by blind investment. This needs to provide practical investment guidance for the majority of small and medium-sized shareholders. The current stock recommendation methods mainly focus on two types. The online stock recommendation method based on stock review and the stock price prediction model based on mathematical analysis have their own advantages, but the defects are quite obvious. The former method can not meet the needs of shareholders' personalized stock recommendation. The application process of the latter method is more complex and difficult to be understood and mastered by investors. The research on the method of stock individualized recommendation has become an important topic at present. In recent years, e-commerce presents a new situation of vigorous development, personalized recommendation method is widely used in the field of e-commerce to target users to recommend goods. Stocks can be regarded as a special kind of goods. Based on this, it is a feasible way to construct the stock personalized recommendation model based on the core idea of the commodity personalized recommendation method. This paper fully studies the literature about the recommendation method at home and abroad. Through the improvement and innovation of the original recommendation method, the stock personalized recommendation model is constructed. The modeling idea is divided into two steps: first. Construct the index system of shareholders' characteristics and use fuzzy clustering method to subdivide the shareholders' groups; Secondly, using graph theory and information maximization retention to improve the original recommendation method, establish a fuzzy clustering based stock personalized recommendation method. After the completion of the modeling process. This paper uses the simulation method to establish the stockholder-stock rating database used in the empirical process, and then uses the programming method to realize the personalized stock recommendation process. Finally. The accuracy and error of recommendation are compared and analyzed with the recommendation method based on collaborative filtering technology and random recommendation method. It is concluded that the personalized stock recommendation method has high recommendation accuracy and is a high performance online stock recommendation method.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51

【参考文献】