كتب heart diseases diagnosis system using multi methods machine learning
اذا لم تجد ما تبحث عنه يمكنك استخدام كلمات أكثر دقة.
Heart Diseases Diagnosis System Using Multi_Methods Machine Learning (كتاب)
Heart disease is a severe illness that can be challenging to diagnose manually. Faster and more precise artificial intelligence models can help diagnose it early. In this work, different detection models were designed to develop a healthy diagnosis system.
The proposed system highlights three objectives: First, an adaptive feature selection technique was proposed using three machine learning methods: Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF). Two feature selection methods, Mutual Information (MI) and Recursive Feature Elimination (RFE), were used to determine the optimal number of features for three binary heart disease tabular datasets from the UCI machine learning repository. Second, various arrhythmia detection models were built using the MIT-BIH arrhythmia database of ECG signals, each using different feature selection and sampling approaches. Machine and deep learning classifiers were combined to create hybrid models that outperform traditional ones. One of these models was the OWSK model, which used a cascading technique that combined the One-Sided Selection (OSS) method-based down-sampling, Wavelet Transform as feature extraction, SVM and K- nearest neighbors algorithms. Finally, a binary classification model based on DenseNet121 was designed to detect cardiomegaly using the CheXpert dataset of chest X-ray images.
The proposed adaptive feature selection technique, with SVM-MI and RF-MI, was the most effective, with the highest classification accuracy and fewer features. On the other hand, the OWSK model achieved 90%, 90%, 93%, 91% and 98%, 98%, 98%, 98%, for weighted accuracy, weighted recall, weighted precision, and weighted F1 score under the inter-patient and intra-patient schemes, respectively. Using a cross-dataset evaluation strategy, the OWSK and cardiomegaly detection models outperformed other models, demonstrating high generalizability and potential for use in medical diagnosis.
