Application Of Knowledge Mining In Time Series Analytical System

With the rapid development of information technology in today's society, people will accumulate a large number of multiple types of time-series data in their daily work. The so-called time-series data is arranged in chronological order, a collection of observations. It has important practical significance to analyze and process with the large number of time-series data and mine with the value of the underlying implication of information. Rough set theory is a powerful mathematical tool brought by Pawlak.Z in 1982, a mathematician in Polish, which could analyze vague and uncertain knowledge. It can effectively deal with incomplete and uncertain knowledge representation and reasoning. This feature makes the rough set theory is very suitable for time-series data analysis and processing.First of all, this paper studies the acquisition approach about the information system of time series, which is the premise of converting to non-sequential information system, but also the foundation of the attribute reduction and rule extraction. Timing information systems can be obtained from the original discrete data also can be converted from real-time timing information system, and then converted timing information system into information system which could use the rough sets to attribute reduction. What's more, we deal the attribute reduction about the decision-making table from the transformation. It is also the foundation of the data mining. This paper minutely describes several processings of the reduction algorithms, and gives examples to verify, and then we get a good algorithm based on set approximation quality. Compared with other algorithms, it is more concise and the effect of reduction is satisfactory. On the other hand, though reduction strategy based on discernibility matrix is widely used but the computation complexity is very large and inefficiency. So in allusion to these problems, this paper has improved an algorithm based on the discernibility matrix, using attribute frequency as heuristic information of choice attribute and using filtering discernibility matrix instead of discernibility matrix to get the attribute frequency,so it can avoid the inefficiency and large computation complexity also can get the satisfactory result of reduction. Finally, this paper studies the methods of rule induction. Because most of the timing signal data in the database is dynamic in actual situation, so the general rule based on static database access method is clearly deficient. This paper presents an incremental rule induction method based on decision tree and rough sets, and compare with the rule induction method which is based on static knowledge system.The experimental results show that the performance of the incremental rule induction method is satisfactory.