Detection of regularity violations of cyclic processes in a temperature monitoring system using patterns form

Periodicity mining is used for predicting trends in time series data. Discovering the rate at which the time series is periodic has always been an obstacle for fully automated periodicity mining. In this paper, a method for detecting the weather temperature series periodicity is proposed. The proposed method, based on DFT, effectively discovered the series periodicity and determined the periodic patterns and their repetition frequencies. Then, the series has been divided into equal time slots based on the pattern repetition frequency. A reference series has been constructed as repetitions for a template pattern, which was constructed from the patterns averages of the original temperature series. The reference series is very useful in temperature series analysis, as the patterns deviations, the future patterns predictions, and the anomalies detections. Experimental results show that the proposed method accurately discovers periodicity rates and periodic patterns.