A topological data analysis based classification method for multiple measurements.

Area of Science:

• Computational Biology
• Data Science
• Machine Learning

Background:

• Machine learning models for analyzing repeated measurements are currently limited.
• Topological Data Analysis (TDA) provides a novel approach to address these limitations.

Purpose of the Study:

• To develop and evaluate a new machine learning classifier for repeated measurements using TDA.
• To compare the performance of the TDA classifier against a standard Support Vector Machine (SVM) model.

Main Methods:

• The TDA classifier samples data, constructs a network graph based on data topology, and applies a cross-validated machine learning model for classification.
• The method was tested on three distinct case studies, including tree species classification, random point processes, and neuron spiking data.

Main Results:

• The TDA classifier achieved 90% accuracy on tree species data, significantly outperforming an SVM model (max 68.7%).
• For random point processes, the TDA classifier reached 96.8% accuracy, vastly outperforming the SVM.
• In neuron spiking data, TDA performance was comparable to SVM in one case (97.8%) but lower in another (79.8% vs. 92.2%).

Conclusions:

• The developed TDA algorithm and software are effective for classifying repeated measurement data, common in biological sciences.
• This approach serves as both an accurate classification tool and a valuable feature selection method.