| The 21 st century is an era of the Internet. The rapid development of the Internet makes people’s life having enormous changes. There is growing accustomed to browse for information and express their views on the network. Virtual social environment of a vast network is a platform for people to express their emotions. People can express their views through the network at any time, which is convenient of community managers to keep abreast of public opinion. Through effective monitoring of network public opinion, social managers can understand the will of people to making the appropriate feedback and correct decision, which is of great significance for social stability.To address the problem, this thesis investigates the network public opinion analysis techniques, and emphatically a variety of technical orientation analysis and experimental researches are carried out. Finally it designs and implements the public opinion tendentious analysis system. The main work is as follows:1. Focus on understanding the tendency of public opinion analysis theory and technology. Study on analysis of the current classification of various existing methods of machine learning.2. This thesis makes use of crawler technology to chime public opinion information from the network, and builds a corpus of public opinion manually for the experiment and analysis.3. Make use of machine learning methods for classification to public opinion propensity analysis for the analysis of public opinion tendency. This article uses active learning and semi-supervised learning algorithm to solve the problem of lack of tagged data. Through experimental analysis and comparison, naive Bayes algorithm and label propagation algorithm has better classification results.4. On the basis of naive Bayes algorithm and label propagation algorithm, this thesis analysis and design the system module, develop and implement a network of public opinion tendency analysis system, the system has good classification results. |
PDF Full Text Request