OPP-Miner: Order-Preserving Sequential Pattern Mining for Time Series.

Area of Science:

• Data Mining
• Time Series Analysis
• Pattern Recognition

Background:

• Traditional sequential pattern mining requires discretizing time series, leading to information loss and ignoring value order.
• Discretization can disrupt time series continuity and obscure important trends.

Purpose of the Study:

• To introduce a novel method, order-preserving sequential pattern (OPP) mining, for analyzing time series data.
• To develop an efficient algorithm, OPP-Miner, for discovering patterns based on the relative order of time series values.

Main Methods:

• OPP mining represents patterns using the inherent order relations of time series values, avoiding discretization.
• The OPP-Miner algorithm utilizes filtration, verification, and pattern fusion strategies for efficient pattern discovery.
• Maximal OPPs are identified to compress the result set.

Main Results:

• OPP-Miner demonstrates efficiency in mining frequent patterns with the same relative order in time series.
• The method successfully discovers similar subsequences within time series data.
• Case studies on COVID-19 data highlight the algorithm's utility in trend identification and clustering improvement.

Conclusions:

• OPP mining offers a robust alternative to traditional methods by preserving time series information.
• The OPP-Miner algorithm is efficient and effective for time series pattern discovery and analysis.
• This approach has practical applications in fields like epidemic analysis and data clustering.