Research On Early Warning Of Web Breaking Events

With the rapid development of information technology, network has become the most important way for people to acquire, publish and share information. The appearance of large amount of web news brings people a great view about what has happened and what is happening all around the world. It's so easy for people to take part in the discussion about something so that people become more and more willing to express their opinions and share interesting things on the Internet. However, the explosive growth of information and the diversification of its transformation can also be along with some bad things. In this paper, we present an algorithm which is used for early warning of web information in order to make it easy for network supervision.Our work can mainly be divided into three parts. First, we come up with a novel online topic detection algorithm which is related to the time factor. Different from traditional algorithm, online topic detection is more sensitive with the influence of time. We fully consider the time factor in this paper. Second, we construct the life span model of topics based on the aging theory, considering many factors to make this model more reality. We train the nutrition transfer factor, the energy decay factor, the clustering threshold; select the energy function and the update method for the key vector of the topic to get better life span curves. Meanwhile, we fully consider the objective factor which could affect the construction of life span models. For instances, the uneven distribution of news stream and the rapid growth when there are just a few documents are challenges for the following work. Finally, we complete the hot topic detection, breaking topic detection and early warning of breaking topics based on the life span models of events and we get high accuracy in experiments. We first get the detection method for hot topics and breaking topics according to the characteristics of these two kinds of topics. Then, we choose different approaches to predict whether a topic could become a breaking topic. The results of experiments indicate that our approach is feasible and effective.