基于图书评论的数据挖掘技术研究

发布时间：2018-03-28 15:54

本文选题：图书评论　切入点：数据挖掘　出处：《北方工业大学》2017年硕士论文

【摘要】：随着互联网的飞速发展,网络信息越来越庞杂,迫切需要使用数据挖掘技术对网络文本进行提取分析,以便得到有用信息。图书评论中包含用户对图书产品属性及购买过程的评价意见,如何从图书评论中高效地提取到有价值的信息成为本文的研究重点。本文以亚马逊和京东两大网站的图书评论为数据来源,研究图书特征提取及情感分析中使用的数据挖掘技术,得到图书评论挖掘结果,从而帮助消费者及生产商作出科学的决策。本文首先通过分析网页的标签规则,抽取网站上的图书评论信息,构建原始评论数据集;利用分词及词性标注技术对该数据集做处理,构建停用词表,对停用词进行过滤,从而形成原始语料库。然后,通过冗余词表处理评论语句。使用Apriori算法,FP-Growth算法以及TF-IDF算法从评论中提取图书特征进行性能比对。以此为基础,改进FP-Growth算法并挖掘图书特征。接着,构建情感词典,对评论进行观点句识别,研究优化SVM特征选择,进行图书评论粗粒度情感挖掘。基于"双向判定法"和情感词典进行评论细粒度情感挖掘,得到具体图书特征的情感极性。最后,使用可视化技术展示图书评论的挖掘结果,并计算用户需求与图书的匹配度,帮助消费者作出购买决策。本文的主要研究成果:第一,通过构建冗余词表进行冗余词的匹配替换,降低了提取频繁项集的冗余度。第二,通过改进FP-Growth算法,将评论长度权重加入算法的支持度计算中,并对提取到的特征进行置信度排序,提高了算法的查全率和准确率。第三,通过优化SVM特征选择,将评论星级作为向量特征加入模型构建过程,提高了情感分析准确度。第四,利用"双向判定法"构建情感关系,实现了图书评论的细粒度情感分析。
[Abstract]:With the rapid development of the Internet, the network information is becoming more and more complex, so it is urgent to use the data mining technology to extract and analyze the network text. In order to get useful information. Book reviews contain users' comments on the attributes of book products and the purchase process, How to efficiently extract valuable information from book reviews has become the focus of this paper. This paper studies the data mining techniques used in book feature extraction and emotional analysis, and obtains the results of book review mining, which can help consumers and manufacturers to make scientific decisions. Extract the book comment information from the website, construct the original comment data set, use participle and part of speech tagging technology to process the data set, construct the stop word table, filter the stop word, and then form the original corpus. The Apriori algorithm and TF-IDF algorithm are used to extract the book features from the comments for performance comparison. Based on this, the FP-Growth algorithm is improved and the book features are mined. Then, the emotion dictionary is constructed. Point sentence recognition of comment, optimization of SVM feature selection, coarse-grained emotion mining of book review, and fine-grained emotion mining of comment based on "bidirectional judgment" and emotion dictionary, Finally, the visual technology is used to show the mining results of book reviews, and to calculate the matching degree between users' needs and books to help consumers make purchase decisions. The main research results of this paper are as follows: first, By constructing redundant vocabulary to match and replace redundant words, the redundancy of extracting frequent itemsets is reduced. Secondly, by improving the FP-Growth algorithm, the weight of comment length is added to the calculation of support degree of the algorithm. Thirdly, by optimizing the SVM feature selection, the comment star is added to the model construction process, which improves the accuracy of emotion analysis. By using the bidirectional judgment method to construct the emotional relationship, the fine granularity emotional analysis of book review is realized.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1

【参考文献】