An Efficient Method for Modeling Nonoccurring Behaviors by Negative Sequential Patterns With Loose Constraints.

Area of Science:

• Data Mining
• Sequential Pattern Analysis
• Machine Learning

Background:

• Temporal Point Processes (TPPs) model sequential discrete events but primarily focus on occurring behaviors.
• Existing Negative Sequential Pattern (NSP) mining methods often impose constraints that limit the comprehensive discovery of non-occurring behaviors (NOBs).

Purpose of the Study:

• To develop an efficient method for NSP mining that leverages TPPs to model both occurring and non-occurring events and behaviors.
• To address the limitations of existing NSP methods by loosening inflexible constraints, enabling a more thorough analysis of NOBs.

Main Methods:

• Introduced a new definition of negative containment based on set theory to accommodate loosened constraints.
• Developed an efficient support calculation method for negative sequences using only information from Positive Sequential Patterns (PSPs), avoiding redundant database scans.
• Proposed a novel and efficient algorithm, NegI-NSP, for identifying highly valuable NSPs.

Main Results:

• The NegI-NSP algorithm efficiently identifies valuable NSPs and discovers more useful NOBs compared to existing methods.
• Experiments on synthetic and real-life datasets validate the efficiency and effectiveness of the proposed approach.
• The method successfully addresses challenges arising from loosened constraints in NSP mining.

Conclusions:

• The proposed method significantly enhances the ability to model and discover non-occurring behaviors within sequential data.
• NegI-NSP offers a more flexible and comprehensive approach to NSP mining, advancing the application of TPPs in sequence analysis.