| Summary: | 碩士 === 國立清華大學 === 資訊系統與應用研究所 === 95 === A method to analyze keywords is proposed in this paper. It measures the weights of keywords in different time slots, and classifies the keywords. According to the features of respecting occurrence of keywords and analyzing Fourier spectrum, which transforms the weights of the keyword in different time slot into frequency, the keywords could be classified. Additionally, we propose a new method, combining full and partial periodicity analysis, to predict the occurrence of keywords, which detects the periodicity of one keyword for prediction.
There are few researchers analyzing keywords by time and utilizing Fourier Transform (FT) for analyzing the result to look for the characteristics of the occurrence of the keyword, and followed by analyzing the features of Fourier spectrum transformed from FT to detect the periodicity of the keywords. The experimental results show that our approach has a good performance in analyzing keywords and predicting the periodicity of keywords. Combining full and partial periodicity analysis, we improve the accuracy of prediction.
|
|---|
