Research On Clustering-based Text Chance Discovery Key Issues

In the real world, an event sequence may include causes, results and some noteworthly events.Typically, the most fundamental reason is hidden, or unknown in some important events. Traditional information processings usually predicted the future possibilites by the frequent event analysis. However, besides the frequent events, there are some rare but important events in it.These events have an important affecting in the future.Chance Discovery has provided a practical theory and method to extract rare important events in observation.Text is an important way of information carrying. Some key issues of Text Chance Discovery is researched based on artificial immune technology and text data oriented in this article.Based on existing Chance Discovery theories and technical methods, the definition of text chance is gived at first, the clustering-based Text Chance Discovery process model is established further. Text Chance Discovery process is a double-helix process by human's experience guid and computer clustering interaction.Secondly, based on the advantages and disadvantages of KeyGraph analysis, a new multi-scan implement model based on KeyGraph is proposed. The detail calculation is realized by using matrix decomposition and so the calculation method is improved. This method effectively enhance the efficiency of the model, decreases the computing data, and reduces the time and space complexity.Thirdly, it is difficult to extract chance quickly and effectively for the number of text data increased rapidly. To solve this problem, text chance discovery pre-processing algorithm is proposed by detailed analysis of the typical chance discovery algorithm pre-processing based on the existing reserch..This algorithm can reduce the complexity, remain the data semantic information and realize the events consistency description. An artifical immune network-based text chance discovery algorithm is proposed. It solves the problem in large scale text data chance discovery by immune network advantages of compressed data size.This extracting algorithm and pre-process algorithm increase the accuracy and effectiveness of text chance extracting, effectivly improve the efficiency and reduce the system storage consumption.Finally, a immune-based multi-agent text chance discovery system framework is proposed, and a dynamic immune network-based response model is established.In this paper, scientific literature as the background, the text chance in sample data set is dicoveried. So that readers can quickly understand the text structure, which verifies the technical advantage of the Text Chance Discovery.