Tsallis statistics enhanced logistic regression for gene expression classification.

Area of Science:

• Computational Biology
• Statistical Modeling
• Machine Learning

Background:

• Classical sigmoid functions have limitations in modeling complex, non-linear data dependencies.
• Existing classification methods may struggle with noise and exhibit instability in real-world datasets.

Purpose of the Study:

• To introduce novel Tsallis statistics-enhanced sigmoid functions for improved classification.
• To evaluate the robustness, noise resistance, and stability of these enhanced functions.

Main Methods:

• Development of two Tsallis statistics-enhanced sigmoid functions using q-deformed exponents (q<1).
• Application of q-deformed classifiers to simulated and four real-world cancer datasets.
• Comparison with traditional methods like Logistic Regression, Support Vector Machine (SVM), and Random Forest.

Main Results:

• Tsallis-enhanced classifiers demonstrated superior robustness, noise resistance, and stability compared to traditional methods in simulated experiments.
• The improved algorithm showed a significantly smaller standard deviation.
• On real cancer datasets, the Tsallis-enhanced method achieved substantial improvements, notably outperforming Logistic Regression with a traditional sigmoid on breast cancer data.

Conclusions:

• Tsallis statistics-enhanced sigmoid functions provide a more flexible and robust fitting method for classification.
• The developed q-deformed classifiers are a reliable solution for complex and noisy data environments, particularly in bioinformatics and medical data analysis.