Discovering shakers from evolving entities via cascading graph inference
Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining
| Hi-index | 0.00 |
This paper studied the representation of time series. The signal and the noise separated by wavelet analysis, the whole data sampling was divided into many continuous intervals, in which the signal was monotone. Each interval was fitted by n-degree polynomial and its eigenvector was made up of the coefficients of the polynomial, its width and Signal Noise Ratio (SNR). The eigenvectors of continuous intervals constituted an eigenvector sequence, which could represent the whole sampling. We analyzed respiratory intensity slice of Chinese astronaut, built the eigenvector sequence and finally found the similar intervals by means of cosine distance of eigenvectors.