面向农资交易的推荐系统设计与实现

发布时间：2018-03-30 06:32

本文选题：农资交易　切入点：推荐系统　出处：《河南大学》2016年硕士论文

【摘要】：为满足用户获得个性化信息的需求,推荐系统得到了快速的发展,被广泛应用到各种领域中,特别是在电子商务领域。在电子商务系统中,利用推荐系统为用户提供相关联或潜在感兴趣的物品,能够针对不同用户的需求提供个性化信息服务,在方便用户获得信息的同时也提升了用户的忠实度。在“互联网+”的时代背景下,农村信息化的进程也开始推进。“新型农村社区电子商务与物流信息服务系统”作为河南省重大专项(新型农村社区信息服务关键技术集成与应用)建设的核心内容之一,面向各类涉农信息,以推荐算法为核心支撑,围绕供求信息进行匹配,为用户提供智能的信息撮合服务,促进农产品流通、加快农业产业化进程、促进农民增收,具有重要的现实意义。基于物品的协同过滤推荐算法是电商系统中应用较为广泛的一种算法,该算法是通过用户对不同物品的评分来评测物品之间的相似度,并基于物品之间的相似度做出推荐。同时,该算法具有不需要领域知识,个性化、自动化程度高,以及随着时间的推移其性能有所提高等优点。但是,在算法的实际应用过程中,需要结合具体应用问题进行调整或改进。具体地:本系统中,在充分考虑物品之间的关联关系的基础上,提供灵活的推荐方式。物品关联关系计算方面:首先创建物品同现矩阵,描述任意两物品之间关系;然后通过构建物品评分矩阵,用来刻画物品之间的关联度,为物品之间的相似度计算提供数据支撑。推荐方式方面:在系统中提供价格优先和距离优先两种推荐方式,当选用不同的方式进行推荐时,分别考虑价格及地域两个属性构建不同的推荐规则,从而使系统灵活稳定的的进行个性化推荐。在分析新型农村社区电子商务与物流信息服务系统的实际需求以及技术路线的基础上,以基于物品的协同过滤推荐算法为核心算法,开展了面向农资交易的推荐系统的设计与实现。本文主要工作如下:1.系统需求分析。首先对系统的总体需求进行分析,讲述对象模型以及系统数据流,在总体需求的基础上,从推荐系统的业务需求以及数据需求两方面细化推荐系统的需求。2.推荐系统架构设计。根据推荐系统的实际需求,为推荐系统设计一种稳定、高效的架构。以推荐系统的业务需求与数据需求为出发点合理的设计数据模型,并以设计的数据模型为前提对基于物品的协同过滤算法进行分析。3.推荐系统设计与实现。创建物品同现矩阵,描述任意两个物品之间的关系。构建物品评分矩阵,分析不同用户对不同物品的评分来评测物品之间的相似性,并计算物品权值为用户进行推荐。对于在不同推荐方式下生成不同推荐规则的算法,使用策略模式的设计思想进行封装,使上层调用者能够统一的处理数据。根据系统中的数据对推荐系统进行验证。
[Abstract]:In order to meet the needs of users to obtain personalized information, recommendation system has been rapidly developed and widely used in various fields, especially in the field of electronic commerce. Using recommendation systems to provide users with associated or potentially interesting items, they can provide personalized information services tailored to the needs of different users. It not only facilitates users' access to information, but also enhances their loyalty. In the context of the "Internet" era, E-commerce and logistics information service system of new rural community is one of the core contents of Henan Province's major project (key technology integration and application of new rural community information service). For all kinds of agricultural information, with the recommendation algorithm as the core support, matching around the information of supply and demand, providing users with intelligent information matchmaking service, promoting the circulation of agricultural products, speeding up the process of agricultural industrialization, promoting farmers' income, The article based collaborative filtering recommendation algorithm is one of the most widely used algorithms in the e-commerce system. The algorithm evaluates the similarity between the items through the users' scores on different items. At the same time, the algorithm has the advantages of no domain knowledge, individuation, high degree of automation and improved performance over time. In the process of practical application of the algorithm, it is necessary to adjust or improve it in combination with the specific application problem. Specifically, in this system, on the basis of fully considering the relationship between items, To provide flexible recommendation methods. In terms of item correlation calculation, we first create an item co-occurrence matrix to describe the relationship between any two items, and then construct an item score matrix to describe the correlation between items. To provide data support for the similarity calculation between items. Recommendation: in the system to provide price-first and distance-first, when the choice of different ways to recommend, Different recommendation rules are constructed based on price and geographical attributes, On the basis of analyzing the actual demand and technical route of the new rural community electronic commerce and logistics information service system, the article takes the collaborative filtering recommendation algorithm based on articles as the core algorithm. The main work of this paper is as follows: 1. System requirement analysis. Firstly, the overall requirements of the system are analyzed, and the object model and system data flow are described. Detail the requirement of recommendation system from two aspects: the business requirement of recommendation system and the requirement of data. 2. The architecture design of recommendation system. According to the actual requirement of recommendation system, a kind of stability is designed for recommendation system. Efficient architecture. A reasonable design data model based on the business and data requirements of the recommendation system. Based on the designed data model, the collaborative filtering algorithm based on articles is analyzed. 3. The design and implementation of the recommendation system. The article co-occurrence matrix is created, the relationship between any two items is described, and the item scoring matrix is constructed. This paper analyzes the different users' scores on different items to evaluate the similarity between the items, and calculates the weight of the items to be recommended by the user. The design idea of policy pattern is used to encapsulate it so that the upper layer callers can deal with the data uniformly and validate the recommendation system according to the data in the system.
【学位授予单位】：河南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.3

【参考文献】