prediction of whether a person is probable to have this disease. Motivated by the growing
mortality of CVD patients every year and the accessibility to a huge amount of patient data
from which to obtain valuable knowledge, we found it useful to use data mining methods
for assisting healthcare professionals in the diagnosis of CVD. The objective of this
research work is not to replace the specialist physician, but to assist the doctor in obtaining
an alternative opinion and its various feasibility in critical situations.
The rest of this paper is organized as follows. Section 2 describes the literature
review. Section 3 presents the proposed approach used for predicting heart disease.
Experimental results are analyzed in Sect. 4 and Conclusion and References are given
in Sect. 5 and 6.
2 Literature Review
In previous studies, researchers expressed their efforts in finding the best model for
predicting cardiovascular disease. In the meantime, various studies give only a glimpse
into predicting heart disease using machine learning techniques and fuzzy logic systems. This section explores the research works that are related to the proposed
approach. A machine learning model has been proposed in [2] by combining five
different algorithms. In fact, the integration of the machine learning model with medical
information systems would be useful to predict the Heart Failure (HF) or any other
disease using the live data collected from patients. A new hybrid approach for heart
disease prediction that combines all techniques into one single algorithm has been
proposed in [4]. The result confirms that accurate diagnosis can be made using a
combined model from all techniques. An “Optimal Multi-Nominal Logistic Regression
(OMLR) algorithm has been proposed in [5] and is used to train the data set for heart
disease. Experiments are conducted on the dataset of UCI heart disease and the results
show 92% accuracy in the detection of heart severity. The Fast Correlation-Based
Feature Selection (FCBF) method has been exploited in [6], to filter redundant features
in order to improve the quality of heart disease classification. Then, the authors performed a classification based on different algorithms such as K-Nearest Neighbour,
Support Vector Machine, Random Forest and a Multilayer Perception optimized by
Particle Swarm Optimization (PSO) combined with Ant Colony Optimization
(ACO) approaches. A predictive model for heart disease diagnosis using a fuzzy rulebased approach with decision tree has been proposed in [7]. In this study, the authors
have obtained the accuracy of 88% which is statistically significant for diagnosing the
heart disease patient and also outperforms some of the existing methods. A new method
namely Hybrid Differential Evolution based Fuzzy Neural Network (HDEFNN) which
can predict the heart disease occurrence fastly and accurately has been proposed in [8].
The performance of this method in terms of accurate diagnosis of heart disease is
attained by improving the initial weight updating of a neural network which is done by
introducing the genetic algorithm. The genetic algorithm can select the most optimal
weight values for the hidden layers of the neural network. A neuro-fuzzy genetic
approach has been proposed in [9], to predict chances of cardiovascular disease. The
proposed approach also helps to make the system more accurate and efficient with the
help of a genetic algorithm.
300
F. Z. Abdeldjouad et al.
mortality of CVD patients every year and the accessibility to a huge amount of patient data
from which to obtain valuable knowledge, we found it useful to use data mining methods
for assisting healthcare professionals in the diagnosis of CVD. The objective of this
research work is not to replace the specialist physician, but to assist the doctor in obtaining
an alternative opinion and its various feasibility in critical situations.
The rest of this paper is organized as follows. Section 2 describes the literature
review. Section 3 presents the proposed approach used for predicting heart disease.
Experimental results are analyzed in Sect. 4 and Conclusion and References are given
in Sect. 5 and 6.
2 Literature Review
In previous studies, researchers expressed their efforts in finding the best model for
predicting cardiovascular disease. In the meantime, various studies give only a glimpse
into predicting heart disease using machine learning techniques and fuzzy logic systems. This section explores the research works that are related to the proposed
approach. A machine learning model has been proposed in [2] by combining five
different algorithms. In fact, the integration of the machine learning model with medical
information systems would be useful to predict the Heart Failure (HF) or any other
disease using the live data collected from patients. A new hybrid approach for heart
disease prediction that combines all techniques into one single algorithm has been
proposed in [4]. The result confirms that accurate diagnosis can be made using a
combined model from all techniques. An “Optimal Multi-Nominal Logistic Regression
(OMLR) algorithm has been proposed in [5] and is used to train the data set for heart
disease. Experiments are conducted on the dataset of UCI heart disease and the results
show 92% accuracy in the detection of heart severity. The Fast Correlation-Based
Feature Selection (FCBF) method has been exploited in [6], to filter redundant features
in order to improve the quality of heart disease classification. Then, the authors performed a classification based on different algorithms such as K-Nearest Neighbour,
Support Vector Machine, Random Forest and a Multilayer Perception optimized by
Particle Swarm Optimization (PSO) combined with Ant Colony Optimization
(ACO) approaches. A predictive model for heart disease diagnosis using a fuzzy rulebased approach with decision tree has been proposed in [7]. In this study, the authors
have obtained the accuracy of 88% which is statistically significant for diagnosing the
heart disease patient and also outperforms some of the existing methods. A new method
namely Hybrid Differential Evolution based Fuzzy Neural Network (HDEFNN) which
can predict the heart disease occurrence fastly and accurately has been proposed in [8].
The performance of this method in terms of accurate diagnosis of heart disease is
attained by improving the initial weight updating of a neural network which is done by
introducing the genetic algorithm. The genetic algorithm can select the most optimal
weight values for the hidden layers of the neural network. A neuro-fuzzy genetic
approach has been proposed in [9], to predict chances of cardiovascular disease. The
proposed approach also helps to make the system more accurate and efficient with the
help of a genetic algorithm.
300
F. Z. Abdeldjouad et al.