基于数据挖掘的时间序列异常事件检测技术研究.pdfVIP

基于数据挖掘的时间序列异常事件检测技术研究 摘 要 dbscan ARMA dbscan ARMA 异常点，时间序列，ARMA ，异常检测 ABSTRACT The detection of abnormal events of time series is of great academic and practical significance for the society today. The new era of anomaly detection of time series is created by the complex features of the continuity and nonlinearity of time series. Time series is one of the special data sequences, and it is widely used in the fields of economy, finance, science and so on. So the anomaly detection of time series is of important status in reality. This paper mainly studies the anomaly detection based on four - point method and the anomaly detection of D SCAN algorithm based on density, and the anomaly detection based on ARMA model. The first is to use the four methods to detect the distribution of abnormal points, which is a simple and conventional method, can remove the obvious abnormal points. The D SCAN algorithm is more complex, is based on the density of a solution, through the neighborhood of the judgment to distinguish between the abnormal points, the result is more accurate. Using ARMA model to deal with the data, judge whether the data is flat can be applied to the choice of the model, then identify the outliers. Key Words Abnormal point, time series, ARMA model, anomaly detection 基于数据挖掘的时间序列异常事件检测研究I 摘要I Abstract II 1 1.1 研究背景1 1.1.1 数据挖掘……………………………………..…………………………………………………………………….1 1.1.2 时间序列数据挖掘……………….………………………………………………………………………….1 1.1.3 异常检测…..……………………………………………………………………………………………………….1 1.2 研究目的…..……………………………………………………………………………