Development Of Data Mining And Analysis System For Smart Watch Based On Online Shopping Reviews

In the era of e-commerce economy,consumers usually post shopping comments on e-commerce platforms to express their views.How to dig out valuable information from a large number of user reviews has an important guiding role for smart watch manufacturers to improve their products.This thesis studies the aspect-level opinion mining and the overall sentiment analysis of reviews for smartwatch online shopping comment mining,and on this basis,develops a smartwatch online shopping comment data mining and analysis system to help companies improve their products more targeted.First,the user needs of a smart watch manufacturer is investigated.Besides,the overall architecture design of the system,the design of various functional modules and database are completed.Secondly,research on aspect-level opinion mining is carried out.The work of attribute clustering is completed,and the attributes of smart watches that users focus on are summarized.A sentiment vocabulary is constructed,and word-level sentiment classification is realized based on dictionary matching.In addition,in view of the characteristics of smart watch online shopping comments,extraction rules combined with dependency syntactic analysis technology are designed to achieve the extraction of attribute-view pairs.Then,the overall sentiment tendency of comments is analyzed.This thesis compares the performance of common text representation methods applied to text sentiment analysis based on traditional machine learning models,and concludes that TF-IDF text representation is the best.On this basis,a general sentiment tendency classification experiment was conducted on traditional machine learning models and deep learning models,and found that deep learning models are not more suitable for text sentiment analysis than traditional machine learning models on the corpus of this thesis.In order to further improve the text sentiment analysis effect of traditional machine learning models,in view of the problems of traditional text discretized representation combined with machine learning models for overall text sentiment analysis,an improved text vectorized representation is proposed and applied to traditional machine learning model.A sentiment analysis comparison experiment with the previous classification scheme was carried out and the results show that the improved text representation can improve the sentiment classification performance of the traditional machine learning model and the classification accuracy rate is higher than that of deep learning model.Finally,the design and development of the data mining and analysis system is completed.The development language is Python,the database management system is My SQL,and the development framework is Django.The main functions implemented include comment collection,text preprocessing,aspect-level opinion mining,overall sentiment analysis,visualization,and system maintenance.After the functional test and performance test of the system,it meets needs of the manufacturer.