Genetik Algoritma Yaklaşımıyla Öznitelik Seçimi Kullanılarak Makine Öğrenmesi Algoritmaları ile Kalp Hastalığı Tahmini

Vatansever, Berna; Aydın, Hakan

doi:10.53525/jster.1005934

Cited by 2 publications

(1 citation statement)

References 40 publications

(27 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the study conducted by Vatansever et al, Decision Tree, Logistic Regression, k-NN, Random Forest, Support Vector Machine and Naive Bayes models were used to predict heart disease with various machine learning algorithms by selecting features with Genetic Algorithm approach. As a result of the study, the highest success rate was obtained with 93.44% in the experiments with the genetic algorithm approach [17]. In the study conducted by Salman and Aksoy, hyperparameter selection was performed with the Random Forest algorithm and then feature selection was performed using the binary particle swarm intelligence method.…”

Section: Related Studies In the Literaturementioning

confidence: 96%

Early-stage heart failure disease prediction with deep learning approach

DEMİR,

BOZKURT,

AYIK

2023

JSR-A

View full text Add to dashboard Cite

Cardiovascular diseases rank the highest among diseases in terms of mortality rate and cause millions of deaths every year. Heart failure is a type of cardiovascular disease and its early diagnosis is extremely important for its prevention. It may be vitally important to understand to what extent which body values, characteristics and factors (age, gender, blood pressure, sugar, etc.) affect this disease and to predict whether the individual will have a possible heart attack in the future. In this study, firstly, the correlation level of the relevant body values with the disease is extracted and in the second stage, a method that predicts heart attack with DNN (Deep Neural Network) and CNN (Convolutional Neural Network) deep learning models is proposed. In the study, 918 observations obtained from the kaggle site were used. Firstly, missing data, categorical data, non-numerical features were checked. Then, outliers were cleaned and the relationship of the features in the dataset with the disease state was revealed by feature engineering operations on the data. Finally, deep neural network models were built and the model was trained and hyperparameter adjustment was performed with GridSearhCV to achieve the highest success rate. As a result of the study, Accuracy, Precision, Recall and F1-Score values were found as 0.9375, 0.9629, 0.9176, 0.9397 for DNN and 0.9312, 0.9512, 0.9176, 0.9340 for CNN respectively. The AUC value calculated from the ROC curve was found to be equal to 0.96 in both deep learning models.

show abstract

Section: Related Studies In the Literaturementioning

confidence: 96%