Improving Alzheimer's Disease Prediction with Different Machine Learning Approaches and Feature Selection Techniques.

Area of Science:

• Computational Neuroscience
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Machine learning (ML) is increasingly vital for healthcare, particularly in disease diagnosis and prediction.
• Alzheimer's disease (AD) diagnosis and prediction remain critical challenges in clinical practice.

Purpose of the Study:

• To develop and compare ML models for predicting Alzheimer's disease (AD).
• To evaluate the effectiveness of feature selection techniques in enhancing ML model performance for AD prediction.

Main Methods:

• Development and comparison of Support Vector Machine (SVM), Random Forest (RF), and Logistic Regression (LR) models.
• Application of feature selection methods: Minimum Redundancy Maximum Relevance (mRMR) and Mutual Information (MI).
• Training and testing models on the OASIS-2 dataset to assess predictive accuracy.

Main Results:

• Logistic Regression (LR) combined with mRMR achieved the highest predictive accuracy of 99.08% for AD.
• Feature selection techniques significantly improved the performance of the ML models.
• Comparative analysis demonstrated varying efficacies among different ML algorithms for AD prediction.

Conclusions:

• ML algorithms, particularly LR with mRMR, are highly effective for accurate Alzheimer's disease prediction.
• Feature selection methods are crucial for optimizing ML model performance in healthcare applications.
• This study highlights the potential of ML in advancing clinical decision-making for neurodegenerative diseases.