基于垂直搜索引擎的农业信息推荐关键技术研究

发布时间：2018-03-18 06:44

本文选题：农业信息推荐　切入点：垂直搜索引擎　出处：《沈阳农业大学》2016年博士论文　论文类型：学位论文

【摘要】：互联网已经成为人们获取信息资源的重要渠道,面对浩如烟海的互联网信息,个性化的信息推荐是未来信息服务的发展方向。另一方面,各级政府和部门投入大量资源建立了涵盖农业科技、畜牧、水产、农垦、农机等领域的信息平台,由于农村地区信息基础建设的缺乏,广大农业生产经营者在信息分析、信息处理等方面能力的匮乏,导致这些对农业生产具有重要指导意义的信息却无法有针对性的传播到农业生产经营者手中。涉农人员仅依靠大众媒体、农业信息机构和口头传播的方式,很难获取到个性化的农业信息服务。该研究的目标是将散布在互联网上的大量农业相关信息进行采集、分析和处理,准确把握涉农用户的意愿和需求,主动将需求信息精准的传播到涉农用户手中,提高农业信息的在农业生产过程中的指导作用和社会经济效益。现有的推荐系统在农业领域的应用主要存在三个问题,一是农业领域信息专注度不够；二是涉农用户兴趣过拟合和冷启动问题；三是现有的信息推荐系统未根据农业的属性特色实现涉农用户个性化的分类和推荐问题。针对以上问题,本研究对农业信息推荐系统的数据源、用户兴趣模型、推荐算法三个重要部件的关键技术进行了深入的研究。主要包括：农业信息采集与分析、用户兴趣模型构建、推荐模型构建和推荐算法改进、软件自主决策机制等关键技术,为个性化农业信息推荐服务的实现提供技术支撑。论文的主要研究工作概括如下：1.通过对搜索引擎功能和搜索效果的比较研究,设计了基于Nutch的农业垂直搜索引擎,实现互联网农业信息的采集、过滤和分析,构建了农业信息推荐资源库。针对垂直搜索在农业领域的应用特点和面临的不足,采用字标注分词技术和参考农业专业术语语料库识别新词的方法改进了搜索引擎的分词模块,实验表明,该分词模块的分词效果与其他分词系统相比,对农业领域文本信息的分词准确度有所提高,结合对种子URL质量的控制,增强农业相关网页的抓取精度和深度。2.针对农业网络资源存在的空间属性表示不统一、显性表达缺失的问题,研究农业领域空间属性信息提取方法,提出了一种借助行政区划本体库对农业领域空间属性的辨别和抽取方法,设计了显性空间属性抽取算法和基于通用搜索引擎的隐性空间属性抽取算法,采用卡方检验的方法解决了隐性空间属性抽取方法中返回空间属性不唯一的问题。两种抽取算法可有效标注网页信息中的空间属性信息,实现用户和项目地域特征的提取,为涉农用户兴趣模型中地域标签的建立和基于地域特征的个性化农业信息推荐模式的实现提供必要的信息。3.采用问卷调查的方法对涉农人员的农业信息需求情况和获取信息方式进行了研究,针对现有的农业信息服务方式无法实现个性化服务的现状,构建了全面反映涉农用户兴趣的模型ATBUIM。选定涉农用户的显式和隐式信息来源,研究了用户背景、浏览行为对用户兴趣度的估算方法和权重,构建了基于互信息和农业领域资源分类标签的贝叶斯网络涉农用户兴趣模型,将农业领域标签间的互信息作为节点条件概率,采用结构学习的方法实现模型的更新和优化。该模型将用户兴趣信息进行加权处理,体现不同类型信息的在模型构建中的比重,更加全面和准确的反映涉农用户的兴趣领域,为实现精准、有效的农业信息推荐算法奠定基础。4.分析和比较了三种推荐算法,针对传统推荐算法存在的冷启动和数据稀疏问题,提出了解决方法和策略,设计了高效的组合推荐算法模型。提出添加特征标签改进算法相似度的方法,解决了传统基于内容推荐算法中新用户无法推荐的问题。针对协同过滤算法中存在的数据稀疏性问题,提出了结合涉农用户的评分、特征因素与农业项目的评分、特征因素的协同过滤算法,算法中目标用户和目标项目的预测评分均为最近邻居综合了评分相似度和特征相似度的结果,加权结合两项预测评分获得最终推荐结果,经实验表明,改进的协同过滤算法在相同数据稀疏度的环境下平均绝对值偏差更小,推荐精度表现更好。针对单推荐算法存在的不足,基于泛函网络提出了一种组合推荐算法,构建了组合推荐模型。实验表明,组合推荐算法计算用户对项目的预测评分更接近用户对项目的实际评分。5.针对信息推荐服务模式在新的网络环境下能够主动调整自身结构、状态和行为的服务需求,提出了一种面向农业领域的软件自主决策机制。基于本体将农业网络信息中的领域知识、消息和服务信息等信息构建模型,设计了面向农业领域知识的思维决策模型AKDM,将环境信息转换成信念、愿望和意图集合,并利用信念-愿望-意图之间的决策推理关系指导Agent完成农业信息推荐行为。分析和实验表明,该机制在农业领域知识和规则的约束下,实现了自主思维决策过程,完成了农业信息的推荐。综上所述,论文对互联网农业信息的有效搜索、涉农用户兴趣模型构建、农业信息精准推荐算法和软件自主决策机制做出的研究,可以为农业领域信息个性化推荐服务的实现提供技术支撑。
[Abstract]:The Internet has become an important channel for people to obtain information resources, facing the multitude of Internet information, personalized information recommendation is the future development direction of information service. On the other hand, governments at all levels and departments to invest a lot of resources covering the establishment of agricultural science and technology, animal husbandry, fishery, agriculture, agricultural machinery and other areas of information platform, due to the lack of information infrastructure in rural areas, the majority of information analysis in agricultural production operators, lack of information processing ability, causes these has important guiding significance for agricultural production information cannot be targeted to spread agricultural production operators hands. Related personnel only rely on mass media, agricultural information institutions and the form of oral communication, it is difficult to get agricultural personalized information service. The aim of this study is a large number of agriculture related information will be spread on the Internet for collection points Analysis and processing, accurate grasp of agricultural user's wishes and needs, the initiative will demand accurate information to spread to the hands of users to improve the agricultural, agricultural information in agricultural production in the process of guiding role and social and economic benefits. The application of the existing recommender systems in the field of agriculture there are three main problems, one is the field of agriculture information focus the two is not enough; related user interest over fitting and cold start problem; three is the information recommendation systems not based on the attribute of characteristic agriculture to realize classification and personalized recommendation problem related to the above problems, this study recommended system data source for agricultural information, user interest model, the key technology of three important components of recommendation algorithm for an in-depth study. Mainly includes: the collection and analysis of agricultural information, build user interest model construction and proposed improved algorithm recommendation model, software The key technology of autonomous decision-making mechanism, to provide technical support for the realization of personalized information recommendation service for agriculture. The main research work is summarized as follows: 1. through the comparative study on the function of search engine and search results, the design of agricultural vertical search engine based on Nutch, realize the Agricultural Internet information collection, filtering and analysis, construction of agriculture recommended information resource. According to the application characteristics of vertical search in the field of agriculture and problems, using the method of word segmentation technology and annotation reference agricultural terminology corpus recognition words improved segmentation module, search engine experiments show that the segmentation module segmentation results compared with other segmentation system, segmentation of the text information in the field of agriculture the accuracy is improved, combined with the control of seed quality of URL, according to agricultural agriculture related web crawler to enhance precision and depth of.2. The spatial attribute existing cyber source said the industry is not unified, the problem of the lack of explicit expression, extraction method of agriculture spatial information, proposes a ontology with administrative identification and extraction method of agricultural field spatial attribute, the dominant design space attributes extraction algorithm based on implicit space and attribute extraction algorithm of general search engine the method of using chi square test to solve the implicit space attribute extraction method in return space property is not the only problem. Two kinds of extraction algorithm can effectively label space attribute information of the web page information extraction of geographical features of users and items, establish agricultural user interest model and method of regional labels to provide the necessary information.3. personalized agricultural information geographical characteristics of the recommended model based on the questionnaire survey on the related personnel of agricultural information demand. Conditions and means of getting information is studied, according to the present situation of agricultural information service currently unable to realize the personalized service, the construction of explicit and implicit information source model ATBUIM. to reflect user interest in the selected agricultural agricultural users, user research background, estimation method and weighted browsing behavior on the user's interest, construction the Bayesian network related user classification and labelling of mutual information and agricultural resources in the field of interest based on the model of agricultural field label between mutual information as the node conditional probability, using the method of structure learning update and optimization model. The model will be weighted to reflect the user interest information, different types of information in the proportion of the model construction. The more comprehensively and accurately reflect the user's interest in agricultural areas, to achieve accurate and effective agricultural information recommendation algorithm to lay the foundation for.4. analysis And comparison of three kinds of recommendation algorithms for cold start and data sparse problem of traditional recommendation algorithm, propose solutions and strategies, design the efficient combination recommendation algorithm model. This method improved algorithm of similarity add feature tags, to solve the traditional content recommendation of new users to recommend the algorithm based on data for. Sparsity in Collaborative filtering algorithm, combining related user ratings, characteristic factor and agricultural project score, collaborative filtering algorithm of feature factors, predictive scoring algorithm of target users and target items are nearest neighbor comprehensive similarity and similarity score results, combined with the weighted score to obtain the final recommendation the two prediction results, the experimental results show that the improved collaborative filtering algorithm on the same data sparsity environment mean absolute deviation is smaller, push The recommended precision performance. Aiming at the shortage of existing single recommendation algorithm, functional network based on a combination of recommendation algorithm is proposed to construct combination recommendation model. Experimental results show that the combination of the recommendation algorithm to calculate the user item rating prediction is closer to the user on the project's actual score for.5. mode recommended service information in the new network environment can take the initiative to adjust its structure, status and behavior of the service demand, put forward a kind of agriculture - oriented software independent decision-making mechanism based on ontology. The agriculture information network in the field of knowledge, model construction and news service information and other information, designed the AKDM thinking decision model for agricultural knowledge, environmental information into belief desire and intention, collection, and use the reasoning relation guide Agent belief desire intention between the completion of the recommended behavior of agricultural information analysis and experimental table. Ming, the mechanism in knowledge and rules in the field of agriculture under the constraints to achieve the independent thinking of the decision-making process, completed the agricultural information recommendation. In summary, the effective search of Internet agricultural information, agricultural user interest model construction, research on agricultural information precision recommendation algorithm and software of autonomous decision-making mechanism, can provide technical support for the realization of personalized recommendation service for agriculture in the field of information.

【学位授予单位】：沈阳农业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.3

【参考文献】