Combining MLC and SVM Classifiers for Learning Based Decision Making: Analysis and Evaluations.

Area of Science:

• Machine Learning
• Computational Biology
• Data Science

Background:

• Maximum Likelihood Classifier (MLC) and Support Vector Machines (SVM) are distinct machine learning algorithms.
• MLC employs Bayesian theory for probabilistic modeling, while SVM is a non-parametric optimization method.
• Recent findings suggest SVM can be equivalent to MLC in probabilistic modeling.

Purpose of the Study:

• To integrate MLC and SVM for improved classification performance.
• To enable probabilistic output for SVMs, facilitating soft decision-making.
• To evaluate the combined SVM-MLC classifier's efficacy.

Main Methods:

• Combining Maximum Likelihood Classifier (MLC) with Support Vector Machines (SVM).
• Utilizing four diverse datasets: sonar, vehicle, breast cancer, and DNA sequences.
• Assessing classifier performance on Gaussian/non-Gaussian and balanced/unbalanced data.

Main Results:

• The combined SVM-MLC classifier demonstrates effectiveness in yielding probabilistic outputs.
• Performance variations were observed based on data characteristics (distribution and balance).
• The integrated approach shows potential for enhanced classification under various conditions.

Conclusions:

• The integration of MLC and SVM offers a novel approach to machine learning classification.
• Probabilistic outputs from SVM enhance decision-making capabilities.
• The SVM-MLC classifier provides a valuable tool for analyzing complex datasets.