Assessing Activity Pattern Similarity with Multidimensional Sequence Alignment based on a Multiobjective Optimization...

Area of Science:

• Computational Social Science
• Bioinformatics
• Optimization Algorithms

Background:

• Human activity pattern analysis is complex due to its multidimensional nature.
• Existing methods struggle with accurately assessing similarity and classifying individuals with similar patterns.
• Developing novel approaches for activity pattern similarity is crucial for understanding human behavior.

Purpose of the Study:

• To present a new methodology for evaluating similarity among individual human activity patterns.
• To conceptualize activity pattern similarity as a multiobjective optimization problem.
• To develop an effective computational approach for classifying individuals based on similar activity patterns.

Main Methods:

• Human activities are coded into multidimensional sequences using sequence alignment.
• Similarity assessment is framed as a multiobjective optimization problem, minimizing alignment costs across all dimensions.
• A multiobjective optimization evolutionary algorithm (MOEA) with specialized operators and local search is employed.

Main Results:

• The proposed multidimensional sequence alignment (MDSA) method effectively evaluates activity pattern similarity.
• The MOEA generates diverse sets of optimal or near-optimal alignment solutions.
• The new method demonstrates superior performance compared to the existing ClustalG method in selected cases.

Conclusions:

• The multiobjective evolutionary algorithm offers an effective approach for assessing human activity pattern similarity.
• This methodology provides a foundation for identifying distinctive groups of individuals with similar activity patterns.
• The study highlights the potential of advanced computational techniques in analyzing complex human behavior data.