Abstract
Anomaly detection is widely used in fields like data processing, intrusion detection, and financial fraud prevention, helping to avoid potential accidents and economic losses. In time series anomaly detection, which deals with numerical sequences over time (e.g., urban temperatures, sales data, stock market trends), the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is an excellent choice. This paper presents an improved anomaly detection algorithm tailored for seasonal time series data. By combining autocorrelation coefficients with the K-means algorithm, precise clustering results down to the date level are provided, subsequently employing the DBSCAN algorithm for detection, the enhanced algorithm is capable of capturing a greater number of local anomalies. Experiment conducted on daily temperature data from Beijing and Sanya in 2023, the enhanced algorithm exhibited a respective increase of 11.6% and 78% in anomaly detection compared to the original algorithm, thus affirming the feasibility of the approach.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call