Applying Feature Selection and Weight Optimization Techniques to Enhance Artificial Neural Network for Heart Disease Diagnosis

Khan, Younas; Qamar, Usman; Asad, Muhammad Usman; Zeb, Babar

doi:10.1007/978-3-030-29516-5_26

Cited by 5 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We used 80% data for training and the rest 20% for the testing. The results of the existing state-of-the-art methods [44][45][46][47][48] and our method are listed in Table VI. From Table VI, we can observe that [47] performed the worst and achieved an accuracy of 88.49%, abdar, [44] performed second best and yielded an accuracy of 93.08% while our method performed the best that yielded an accuracy of 93.44% on Z-Alizadeh sani dataset.…”

Section: Performance Comparison With Other Methodsmentioning

confidence: 99%

“…Research community explored various cardiovascular disease prediction systems [24-29], [44][45][46][47][48], [51,54] based on machine learning algorithms. In [24], different models were employed for the prediction of heart diseases i.e., approximated L0-norm SVM strategies, recursive characteristic elimination, and standard SVM, but able to achieve an accuracy of 76.8% only due to using unnecessary features.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, the research community proposed numerous machine learning techniques [13][14][15][16][17][18][19][20][21] in diagnosing various diseases. Classi cation in medical sciences is the most important and popular decision-making tool for the diagnosis of diseases.Research community explored various cardiovascular disease prediction systems [24-29], [44][45][46][47][48], [51,54] based on machine learning algorithms. In [24], different models were employed for the prediction of heart diseases i.e., approximated L0-norm SVM strategies, recursive characteristic elimination, and standard SVM, but able to achieve an accuracy of 76.8% only due to using unnecessary features.…”

mentioning

confidence: 99%

“…An accuracy of 98.2% was achieved using mapping-based cuttle sh optimization algorithm for features selection. In [47], four distinctive features selection methods i.e., information gain, gini index, principal component analysis, and weights by SVM were used to select optimal subset of features whereas, four optimization techniques i.e., evaluation strategy, particle swarm optimization, forward and backward weights optimization were employed to enhance the neural network using 10-fold cross validation technique.…”

mentioning

confidence: 99%

“…Even, if the literature signi es much improvement to detect the heart disease, the problem is far being solved. Still, there exists certain limitations of the existing state-of-the-art heart attack prediction methods [44][45][46][47][48][49][50][51][52][53][54][55][56][57] i.e., low accuracy, expensive treatment, unable to better handle missing values and high dimensional datasets, undiscovered valuable information from clinical datasets, usage of irrelevant variables which degrades the system performance, using small datasets for training purpose that are unreliable for heart disease prediction, and incapable of extracting prominent features from diverse datasets. Therefore, to achieve the maximum capability of machine learning and deep learning algorithms, it is necessary to pre-process the clinical data to achieve better classi cation accuracy because clinical datasets are irregular and uncertain.…”

mentioning

confidence: 99%

See 4 more Smart Citations

CNN-CardioAssistant: Deep Convolutional Neural Network and Recursive Feature Elimination Method for Heart Disease Detection

Hassan

Rahman

Javed

et al. 2022

Preprint

View full text Add to dashboard Cite

In recent times, we have seen an exponential rise in different chronic diseases due to our unhealthy lifestyles. Cardio disease is the most common and life-threatening among all diseases, which contributes to a very high mortality rate. Accurate detection of cardio disease at an early stage is vital to save the lives of people. Most of the existing cardiovascular disease detection systems suffer from lower performance and efficiency due to redundant attributes, dimensionality curse, imbalance, and noisy datasets. In this work, we proposed a novel convolutional neural networks-based system (CNN-cardioAssistant) that predicts cardiovascular disease in patients. Recursive feature selection (RFE) is employed to select more prominent features from the clinical data of cardio patients. The selected features are then used to train the proposed CNN-CardioAssistant as well as 11 different classifiers i.e., support vector machine (SVM), Random Forest, Decision Tree, logistic regression, Naïve Bayes, K-Nearest Neighbor, XGboost, Multi-layer Perceptron, Gaussian process classifier, Adaboost, and Quadratic discriminant analysis separately for cardio disease prediction. We compared the results of all the methods on three subsets of features i.e., 6,8, and 15 for each dataset. The features selection method provides optimal subsets of features that can reliably be used to predict cardiovascular disease with the highest accuracy. Experimental results on three different cardio datasets i.e., Public Health, Framingham, and Z-Alizadeh sani clearly demonstrate that the proposed CNN-CardioAssistant system has superior performance against the existing state-of-the-art methods.

show abstract