Efficient Mining of Weighted Flexible Periodic Patterns in Time Series Databases using a Period-based Algorithm with Multiple Join Policies

Periodic pattern mining in time series databases is crucial for predicting future trends. The L-PPM algorithm is a list structure of the flexible periodic pattern that can increase the flexibility of the mining result. However, it will generate an invalid candidate, like the pattern with the last symbol = '*'. To solve this, we propose a Period-based algorithm with a PDH table. It allows for parallel processing and avoids invalid candidates. Performance studies show its efficiency over the L-PPM algorithm.