A Feature Selection Approach for Enhancing the Cardiotocography Classification Performance

Data Mining is an interdisciplinary field of study for knowledge discovery from the large data sets. Collection of data, preprocessing, applying intelligent data mining techniques and interpretation are variou stages of data mining process. Out of which preprocessing is the critical part in data mining process. Noise, mislabelled data, imbalanced class, feature selection are some of the issues involved in preprocessing. In this research feature selection methodology has been considered for improving classification result for predicting Cardiotocography class. We have considered Correlation based Feature Selection(CFS), Symmetrical Uncertainty, ReliefF, Information Gain, Chi- Square feature selection methods and four different types of classifiers namely Jrip (Rule based), J48 (Tree based), NB (Bayes Learner), KNN (Lazy Learner). Proposed method has recorded some better result than considering all the features.