Artificial liver classifier: a new alternative to conventional machine learning models.

Area of Science:

• Machine Learning
• Computational Biology
• Artificial Intelligence

Background:

• Supervised machine learning classifiers often encounter challenges with performance, accuracy, and overfitting.
• Developing robust and efficient classification models remains a key area of research.

Purpose of the Study:

• Introduce the Artificial Liver Classifier (ALC), a novel supervised learning model.
• Address limitations of current classifiers, focusing on simplicity, speed, and overfitting reduction.
• Investigate the efficacy of the ALC for multi-class classification problems.

Main Methods:

• The Artificial Liver Classifier (ALC) is a new supervised learning model inspired by the human liver's detoxification process.
• Parameter optimization for the ALC is performed using an improved FOX (IFOX) optimization algorithm.
• The ALC model was evaluated on five benchmark datasets: Iris Flower, Breast Cancer Wisconsin, Wine, Voice Gender, and MNIST.

Main Results:

• The ALC achieved high accuracy, reaching 100% on the Iris dataset and 99.12% on the Breast Cancer dataset.
• ALC outperformed established classifiers like logistic regression, multilayer perceptron, support vector machine, and XGBoost on specific datasets.
• Across all tested datasets, the ALC demonstrated smaller generalization gaps and lower loss values compared to conventional methods.

Conclusions:

• Biologically inspired models offer a promising direction for developing efficient machine learning classifiers.
• The Artificial Liver Classifier (ALC) presents a viable and effective approach for multi-class classification tasks.
• The study opens new avenues for innovation by integrating biological principles into machine learning model design.