A Study on Customers' Sentiment Analysis Based on Big Data Using Twitter Data

Abstract

This paper focuses on mining the value of customers emotional behaviors using Twitter data. Using Apache Flume to collect tweets data from Twitter. 192,390 tweets were collected. Then the natural language processing (NLP) technology has been used to divide and filter tweets for customers' emotional behaviors analysis. We picked five main hot topics among these tweets. Choose one of the hot topics HUAWEI honors 9 for sentiment analysis (SA). Compared with Native Bayes, Maximum Entropy Classifier. Decision Tree Classifier is the most effective classification method for our data sets. According to our experiment, the result shows that 45% of customers are satisfied with HUAWEI honors 9, but there is still having 36% of customers unsatisfied with it. Specifically, in the field of battery, game and stand-by power consumption, it needs a great of an improvement.