Abnormal Pre-Detection Algorithm Based on Time Series

For these data sets whose attributes changing with time, it is more practical to avoid the occurrence of anomalies in time. Based on the abnormal detection, an improved density-based algorithm Pre-ADT (Previous-Abnormal Detection) was proposed, which predicts the abnormality of the data at the next time point, and avoid the generation of outliers in time. Firstly, the algorithm improves the DBSCAN(Density-Based Spatial Clustering of Applications with Noise) algorithm for density clustering, and finds the point in time at which an abnormality may occur by increasing the weight of the abnormal fluctuation attribute point, then it expresses the abnormal degree according to the local outlier factor of the LOF(Local Outlier Factor) algorithm. For the data points whose abnormal tendency is more and more obvious, the algorithm uses the linear regression algorithm to predict the next point, and the predicted values rise exponentially, for the intermittent abnormal fluctuations, it also retains its certain influence, and ensures the sensitivity of the algorithm to the abnormal points. The experimental results show that this algorithm can predict the occurrence of abnormal conditions in a timely and effective manner.