Based On Machine Learning Research And Implementation Of Network Public Opinion Collection Technology

With the rapid development of Internet technology, network speech platform of increasingly prominent, and the network false, violence, negative public opinion impact on social stability and national security threat has become increasingly serious. So the effective acquisition of network public opinion in preventing the spread of bad information ensure public security has important significance.This thesis focuses on analyzed the key technologies and improved collection system network public opinion analysis: text clustering. This thesis designs and realizes a network public opinion collection system.1. In this thesis, the Single- Pass of text clustering algorithm is improved. As the network public opinion collecting technology based on machine learning, unsupervised machine learning text clustering algorithm is its core. This thesis designed a kind of Single-Pass algorithm based on double threshold, through the establishment of the middle class status of the cluster center vector offset specification to reduce the dependence of the intensity of the input sequence. Experiments show that the improvement on the performance of the text clustering has large improvement.2. This thesis improves the DOM tree to improve the text-based extraction methods. This approach optimizes the traditional text extraction method based on DOM tree by adding Chinese characters and distribution of the ratio of non-link text. This method improves the body of public opinion acquisition system extraction accuracy.3. This thesis build a network of public opinion collection system architecture based on machine learning, designed and implemented a prototype system and the whole of its core modules and systems were tested.