Intelligence System for Diagnosis Level of Coronary Heart Disease with K-Star Algorithm

Abstract

ObjectivesCoronary heart disease is the leading cause of death worldwide, and it is important to diagnose the level of the disease. Intelligence systems for diagnosis proved can be used to support diagnosis of the disease. Unfortunately, most of the data available between the level/type of coronary heart disease is unbalanced. As a result system performance is low.MethodsThis paper proposes an intelligence systems for the diagnosis of the level of coronary heart disease taking into account the problem of data imbalance. The first stage of this research was preprocessing, which included resampled non-stratified random sampling (R), the synthetic minority over-sampling technique (SMOTE), clean data out of range attribute (COR), and remove duplicate (RD). The second step was the sharing of data for training and testing using a k-fold cross-validation model and training multiclass classification by the K-star algorithm. The third step was performance evaluation. The proposed system was evaluated using the performance parameters of sensitivity, specificity, positive prediction value (PPV), negative prediction value (NPV), area under the curve (AUC) and F-measure.ResultsThe results showed that the proposed system provides an average performance with sensitivity of 80.1%, specificity of 95%, PPV of 80.1%, NPV of 95%, AUC of 87.5%, and F-measure of 80.1%. Performance of the system without consideration of data imbalance provide showed sensitivity of 53.1%, specificity of 88,3%, PPV of 53.1%, NPV of 88.3%, AUC of 70.7%, and F-measure of 53.1%.ConclusionsBased on these results it can be concluded that the proposed system is able to deliver good performance in the category of classification.

Hybrid System of Tiered Multivariate Analysis and Artificial Neural Network for Coronary Heart Disease Diagnosis

“…Furthermore Mokeddem et.al [24] by using a wrapper feature selection, which is implemented by genetic algorithm and C4.5, produce a number of attributes that are less than the proposed system, but weaknesses attributes generated, has two attributes costly in examination. The next comparison with research conducted Muthukruppan & Er [7], Abdar et.al [27], Wiharto et.al [28] and Subanya & Rajalaxmi [3]. These studies are able to provide higher accuracy performance as compared to the proposed system.…”

“…Unfortunately, the high accuracy should still require costly attribute, namely scintigraphy examination and flouroscopy. In addition the number of attributes required in research Muthukruppan & Er [7] and Wiharto et.al [28] to produce an accuracy above 90% require a relatively large number of attributes compared to the proposed system. While the research conducted Abdar et.al [27], using logistic regression attribute is able to reduce from 13 to 6 attributes, and with C5.0 algorithms capable of generating an accuracy above 90%.…”

“…Tiered logistic regression method using all the attributes is also capable of providing AUC value of 0.915, the value is still lower than the proposed system. [5] GA Wrapper + C4.5 3,6,12,13 78,54% Mokaddem et.al [5] GA Wrapper + MLP-NN 1,2,3,4, 11,12,13 79,86% Khemphila & Boonjing [26] Neural Network 1,2,3,4,5,6,7,8,9,10,11,12,13 80,20% Abdar et.al [27] Logistic Regression + Neural Network 2,3,4,9,12,13 80,23% Bashir et.al [25] Majority Voting (Naive Bayesian, Decision Tree, SVM) [27] Logistic Regression + kNN 2,3,4,9,12,13 88,37% Wiharto et.al [28] k-start 1,2,3,4,5,6,7,8,9,10,11,12,13 92,02% Abdar et.al [27] Logistic Regression + C5.0 2,3,4,9,12,13 93,02% Muthukaruppan & Er [7] Particle Swarm Optimization+ Fuzzy Inference Subsequent research conducted by Arjenaki et.al [9], which combines genetic algorithm with naive Bayesian. Fitness function used in the genetic algorithm is a function of the examination fee.…”

Prediction of G Protein-Coupled Receptors with SVM-Prot Features and Random Forest

“…(5) The training and test datasets were successively imported into weka data mining package ( http://www.cs.waikato.ac.nz/ml/weka/ ), a machine-learning workbench. In weka, the training datasets were filtered with the synthetic minority oversampling technique (SMOTE) [ 45 , 46 ] and changed the positive samples from 100 percent into 300 percent to overcome the highly imbalanced property of positive and negative cases; after preprocessing with SMOTE technique the two-group data kept an amount equilibrium, and the vector data were classified automatically via visualization analysis [ 47 ]. Based on the optimal features with some preliminary trials, we finally chose a Random Forest (RF) [ 48 ] module and “use training set” item on test options as classifier for training dataset, while for test dataset we chose “supplied test set” item on test options to predict the samples as GPCRs or non-GPCRs: that is, the prediction module using the results of the just training set to distinguish the two classes.…”

Identification of DEP domain-containing proteins by a machine learning method and experimental analysis of their expression in human HCC tissues

“…In weka, we filtered the vector data with the synthetic minority over-sampling technique(SMOTE)777879 and changed the positive instances from the 100% into 700% to overcome the highly imbalanced data. the vector data were automatically classified by visualization and cross-validation analysis808182838485.…”