Stock time series visualization based on data point importance

Time series visualization is a fundamental task in most financial applications. A framework that can reduce the dimensionality of the time series data, sufficiently accurate so that it can capture the actual shape of the time series but, at the same time, salient points will not be smoothed out can take advantage of charting the raw time series data. On the other hand, it is preferable that the representation framework can handle the multi-resolution problem rather than reduce the dimension to a fixed level only. In this paper, a framework that represents and visualizes time series data based on data point importance is proposed. Furthermore, discovering frequently appearing and surprising patterns are non-trivial tasks in financial applications. A method for discovering patterns across different resolutions is proposed. The proposed method is based on a modified version of VizTree. By converting the time series to symbol string based on data point importance, the potential patterns with different lengths can be encoded in the VizTree for visual pattern discovery while the important points and the overall shape of the time series patterns can be preserved even under a high compression ratio. Various experiments were conducted to evaluate the performance of the proposed framework. One may find it particularly attractive in financial applications like stock data analysis.