Research On The Model And Algorithm Of Series Data Mining

In several years, data mining has spread over the world and become the one of thehottest spot of research in IT field. Researchers from various fields show great enthusiasmfor data mining and are willing to devote themselves to the new field because data miningsyncretizes artificial intelligence,statistics,database and machine study. At the beginning ofthe new century, it is a large-scale amalgamation of subjects. The objective of data mining is to find out the interesting model from a great deal ofdata in order to figure out the regulation, improve the effect and predict the future. Thecore of data mining is of great challenge. Due to the data's timing and time feature, thingssuch as web mining, finance data mining, e-business and market research would deal withserial data unavoidably. The serial data mining is birth to meet the sort of needs. Thedifference between serial data mining and other branches is that the serial data miningdepends on model much more than the latter. As we all know, time serial analyse is famousfor its complexity and abstract,especially its relating to one difficult problem— outliermining. To solve the time series is still a hard work even in the field of statistics. There isno perfect solution so far. Financial data mining is another difficult problem. Thoughfinancial math has given many methods and abundant theory, some of these methods areabstractive and less effective in practice. So popularizing these methods is not meaningful.As IT technology, data mining should be characterized by simplicity and speediness. It isjust the aim of the paper to figure out the tangible method and theory system both of whichare based on arithmetic and models. In the chapter one, we introduce concepts, basic knowledge, status quo at home andabroad, linear model in statistics and its apply in serial data mining. In the chapter two, we aim to the parameter estimating of linear model and achievemany new profound results. In the chapter three, we mainly put emphasis on data visualization,cluster of serialdata, linear model and outlier mining. We apply these methods and theory to stock datamining and achieve satisfactory results. In the chapter four, we focus on the prediction of serial data mining and also getstrong support from stock demonstration analyse. In the chapter five, we research data repair and interpolation such as EM- arithmeticin order to solve data lack that has often happened. In addition, we research forward search II重庆大学博士学位论文 英文摘要 procedure for the further step and reach the goal of saving time and avoiding mistake byintroducing cluster pretreatment to forward search. The field of serial data mining is so wide that we cannot cover all of it. At the end ofthe paper, we list many of problem unsolved and predict the development of the brand.