Pattern Mining in Sparse Temporal Domains, an Interpolation Approach

Weblog systems, mobile phone companies or GPS devices collect large amounts of personalized data including temporal, positional and textual information. Patterns extracted from such data can give insight in the behavior and mood of people. These patterns are often imprecise due to sparseness in the data. We propose an interpolation technique that augments local patterns with elements that seem to be locally unimportant but with global information they are interesting.