Analysis Of Commodity Reviews Based On Text Mining

With the rapid development of e-commerce,there are countless comments.The subjective evaluation made by the customer on the products and services purchased is a double-edged sword.Good evaluations will bring invisible advertising effects to merchants and it is the important tool to enhance their competition ability.Negative evaluations can seriously affect customers’ purchase intentions and cause losses to businesses.Therefore,analyzing the evaluations of goods can provide a strong basis for merchants to discover products’ weakness and improve the quality and service attitude.And then.they can stop loss in time and establish a good brand image.However,the number of online reviews is increasing day by day,and it is difficult to organize and analyze comments by manual means.It is necessary to study the text mining methods.This paper takes the comments of mobile phone as the research object.Firstly,a word segmentation dictionary and a stop word dictionary are built,and then use the natural language processing technology to deal with the comments.And then using vector space model to convert text data into word frequency matrix,the text can be transformed into traditional data structure.In order to find the consumer’s view of this mobile phone’s functions,the good and bad comments are divided and arrange them in descending order of word frequency.And select the mobile phone’s functions from nouns of the high-frequency words.This is the mobile phone’s functions that the consumer focuses on.The results show that businesses need to improve their service and manage price.The screen,battery power consumption,and signal also need to be improved.The appearance,color,cost performance,quality,and speed are recognized.This paper uses Naive Bayes and Random Forest to classify the text.Both methods are applied to the test to check out the classification effect after the classification model is established for the training set.The experimental results show that the accuracy of the two classification methods is higher than 85%,and the recall rate is higher than 90%.When the number of term is relatively small,the Naive Bayes’ effect is better,and the Random Forest is better when the number of term is larger.This paper proves the validity of these two methods for the emotional classification.