Application Of Data Mining Based On Rough Set Theory In Time Series Analysis System

Time-series data is an important class of complex data, which exists in almost all areas of our society, such as medicine, weather forecasting, stock prediction, network intrusion detection and moving-object simulation. How to analysis these time-series data effectively, research the information which hidden in the data and help people understand things correctly, and thus make scientific decision-making, because of these data mining has caused widespread concern. Time-series data, including two types one is multivariate time series, the other is one variable time series data. The studies of one variable time series is much, and has gradually formed mature theories and methods; but the structure of multivariate time series is more complex than one variable, and the existing theories and methods are not perfect. Multivariate time-series data are largely collected in the areas of financial, medical, process control and so on. Not all of these time series are meaningful. Some of time-series have certain correlations, to find such sequence for forecasting will reduce the difficulty of analysis and increase the accuracy. So to study the classification of multivariate time-series is great significance.First of all, this paper has studied the methods of how to get the attributes of time series, including the basic research methods:derivation of the time-series data period, the derivation of the time-series Fourier coefficient which used least square method, correlation analysis and so on. They are not only the premise of the time-series analysis, but also the basic of using rough set theory. Secondly, the paper introduced the theoretical concepts of rough set theory and the ideas of classification based on attribute. Again, this paper analyzed the time complexity of traditional classification for the multiple time-series data. For the drawback of the highly complexity a new improved algorithm is proposed, which are the classification algorithm based on rough sets, and also particularly analysis the complexity of the new algorithm. Comparing the time complexity of two algorithms I found the new algorithm with features of fewer iterations and faster classification, so greatly improved the classification efficiency. Finally, I used a set of experimental data to verify the accuracy of the new algorithm, the results show that the proposed method and the original classification algorithm have the same result, so verify the new method is feasible.