dc.description.abstract | System log data includes the records of system users’ operation. By means of analyzing log data, we can get much valuable information about system efficiency, users’ habitual behaviors and interests, etc. This information is useful to realize system users more, even to help set up proper strategies.
Many systems and technologies of analyzing log records are based on viewpoint of data. However, who needs to analyze data is who should decide the content of data for analysis. Analyzing log records based on viewpoint of data, if the data users required doesn’t exist in systems, usually increases users’ loads by data rearrangement and recalculation.
Hence we design an on-line analytical processing system based on viewpoints of system users to analyze data. First, by operating the split process, users can define various kinds of data features to give more sensible meanings for cube data. In this part, we offer several split functions. Users can model data features by these functions and increase credibility of data feature definitions. And then operating replacing process, users are able to construct feature space to obtain feature-related data. Moreover, the system provides Boolean operation for users. So users can consider multiple conditions to define features or construct feature space. Through data features, users combine their viewpoints with cube data and get required data directly.
After gaining data, we want to do some analysis. The analytical results will contain these feature definitions. Besides data mining technologies, we also adopt statistical methods for data analysis. People have affirmed the test ability of statistics. By using some statistics methods, users will put more trust in analytical results. | en_US |