| Summary: | 碩士 === 國立臺灣科技大學 === 電子工程系 === 91 === In recent years, there are many researches on mining the similarity from time-series data, most of than are based on which adopt the method of reducing the dimension of the data by Discrete Fourier Transform (DFT) and Discrete Wavelet Transform (DWT) and then sorting out the similarity according to their Euclid distance. Because of they are complicate and time-consuming in the reduction of data dimensions and calculation of distance, we propose to use matched filter and correlator technology, a method that has the advantages of speeding up the similarity mining process, and increasing the accuracy.
In this thesis, the time-series data will be treated as signals. We identify the similarity between the two signals from the energy they contain by applying the matched filter and correlator. In addition to the general ways of querying the information on the similarity, we also develop an the auto-searching engine for establishing the similarity index from the time-series data, so that it would be convenient for the subsequent work on periodic mining and the relationship analysis. Finally, we adopt the air quality monitoring dataset collected from the Environmental Protection Administration Government of the Republic of China as our sample to conduct the empirical experiment to verify the feasibility of the proposed method.
|
|---|
