Optimal feature selection using a modified differential evolution algorithm and its effectiveness for prediction of heart disease

• Published in: Computers in biology and medicine Vol. 90; pp. 125 - 136
• Main Authors: Vivekanandan, T., Sriman Narayana Iyengar, N. Ch
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.11.2017; Elsevier Limited
• Subjects: Automatic Data Processing - methods; Big data; Cardiovascular disease; Coronary artery disease; Data processing; Datasets; Differential evolution; e-Healthcare; Electronic Health Records; Evolutionary algorithms; Feature selection; Feed-forward neural network; Fuzzy AHP; Health care; Heart; Heart diseases; Heart Diseases - diagnosis; Humans; Mathematical models; Medical records; Neural networks; Neural Networks (Computer); Optimization; Strategy; Feature selection; Differential evolution; Big data; Feed-forward neural network; e-Healthcare; Fuzzy AHP; Optimization
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2017.09.011

Enormous data growth in multiple domains has posed a great challenge for data processing and analysis techniques. In particular, the traditional record maintenance strategy has been replaced in the healthcare system. It is vital to develop a model that is able to handle the huge amount of e-healthcare data efficiently. In this paper, the challenging tasks of selecting critical features from the enormous set of available features and diagnosing heart disease are carried out. Feature selection is one of the most widely used pre-processing steps in classification problems. A modified differential evolution (DE) algorithm is used to perform feature selection for cardiovascular disease and optimization of selected features. Of the 10 available strategies for the traditional DE algorithm, the seventh strategy, which is represented by DE/rand/2/exp, is considered for comparative study. The performance analysis of the developed modified DE strategy is given in this paper. With the selected critical features, prediction of heart disease is carried out using fuzzy AHP and a feed-forward neural network. Various performance measures of integrating the modified differential evolution algorithm with fuzzy AHP and a feed-forward neural network in the prediction of heart disease are evaluated in this paper. The accuracy of the proposed hybrid model is 83%, which is higher than that of some other existing models. In addition, the prediction time of the proposed hybrid model is also evaluated and has shown promising results. Part 1 : Modified DE Based Optimization of Feature selection for Heart Disease Data. [Display omitted] •e-Healthcare data predicting cardio vascular disease has been taken for analysis.•Medical dataset from UCI repository, which consist of 300 individuals' Heart disease data, has been taken for study and the available 13 attributes have been reduced to 9 critical attributes, thus efficiently huge data has been handled using modified DE strategy.•The output of Modified DE is fed in to an integrated model of Fuzzy AHP with Feed Forward Neural Network.•The output of the integrated model predicts the presence or absence of heart disease.•The performances of Modified DE and the prediction of heart disease have been analysed.