LCSS-Based Algorithm for Computing Multivariate Data Set Similarity: A Case Study of Real-Time WSN Data.

Area of Science:

• Computer Science
• Data Analysis
• Algorithm Development

Background:

• Multivariate data analysis is crucial in fields like wireless sensor networks (WSNs) and DNA analysis.
• Existing similarity index computation methods, particularly dynamic programming, face efficiency challenges with large multivariate datasets.
• Classical approaches struggle with high/low similarity indexes and specific data types like sensor data.

Purpose of the Study:

• To propose an efficient algorithm for measuring similarity indexes in multivariate data sets.
• To utilize a non-metric-based methodology, specifically the longest common subsequence (LCS) technique.
• To overcome the limitations of existing dynamic programming methods in terms of efficiency and applicability.

Main Methods:

• Developed a novel, efficient algorithm for non-metric similarity index computation.
• Employed the longest common subsequence (LCS) approach as the core non-metric methodology.
• Evaluated the algorithm's performance against classical dynamic programming algorithms.

Main Results:

• The proposed algorithm demonstrates superior performance on various multivariate datasets compared to dynamic programming methods.
• Significant efficiency gains were observed, with the new algorithm being approximately 39.9% faster in computational time.
• The algorithm proved effective on both benchmark and real-world dynamic multivariate data from a WSN.

Conclusions:

• The novel non-metric algorithm provides a more efficient and robust solution for multivariate data similarity analysis.
• This approach offers a practical improvement over traditional dynamic programming techniques, especially for WSN and DNA data.
• The findings suggest broader applicability in domains requiring efficient similarity computation for complex datasets.