SRADHO: statistical reduction approach with deep hyper optimization for disease classification using artificial...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Neuroscience

Background:

• Brain-related diseases significantly impact cognitive function, necessitating accurate and early diagnostic tools.
• Artificial neural networks show promise in disease prediction but face challenges like overfitting and performance degradation.
• Efficient feature selection and hyperparameter optimization are crucial for robust medical data analysis.

Purpose of the Study:

• To present a novel Statistical Reduction Approach with Deep Hyper Optimization (SRADHO) for enhanced disease classification.
• To address overfitting, underfitting, and reduce computational time in medical data analysis.
• To improve the accuracy and efficiency of brain disease diagnosis using AI.

Main Methods:

• Developed SRADHO, combining deep learning with hyperparameter tuning for automatic feature identification and dimensionality reduction.
• Utilized Bayesian optimization within SRADHO to calibrate model weights, biases, and select optimal hyperparameters.
• Experimented on three benchmark datasets using various classifiers including logistic regression, decision tree, random forest, k-NN, SVM, and Naïve Bayes.

Main Results:

• The SRADHO algorithm achieved high performance metrics: 98.2% accuracy, 97.2% precision, 98.3% recall, and 98.1% F1-Score.
• Demonstrated a significantly reduced error rate of 0.3% and an execution time of only 12 seconds.
• Effectively optimized feature selection and model parameters, outperforming traditional methods.

Conclusions:

• SRADHO offers a statistically robust and computationally efficient approach for brain disease classification.
• The technique successfully mitigates common deep learning challenges, leading to superior diagnostic accuracy.
• SRADHO holds significant potential for advancing early and accurate disease diagnosis in clinical practice.