An efficient hybrid approach based on SVM and Binary ACO for feature Selection O. KADRI * , L.H. MOUSS * , F. MERAH ** , A. ABDELHADI * & M. D. MOUSS * * Laboratory Automation & Production Engineering University of Batna 1, Rue Chahid Boukhlouf 05000 Batna, ALGÉRIE ** Department of Mathematics University of Khenchela Route de Batna BP:1252, El Houria, 40004 Khenchela ALGÉRIE ouahabk@yahoo.fr, merahfateh@yahoo.fr, hayet_mouss@yahoo.fr, abdelhadi.adel@yahoo.fr & djmouss@yahoo.fr Abstract One of the most important techniques in data pre- processing for classification is feature selection. In this paper, we propose a novel hybrid algorithm for feature selection based on a binary ant colony and SVM. The final subset selection is attained through the elimination of the features that produce noise or, are strictly correlated with other already selected features. Our algorithm can improve classification accuracy with a small and appropriate feature subset. Proposed algorithm is easily implemented and because of use of a simple filter in that, its computational complexity is very low. The performance of the proposed algorithm is evaluated through a real Rotary Cement kiln dataset. The results show that our algorithm outperforms existing algorithms. Keywords Binary Ant Colony algorithm, Support Vector Machine, feature selection, classification. 1. Introduction Our work falls under the Condition monitoring and diagnosis of industrial system which is an important field of engineering study (in our case is a Rotary Cement kiln, see fig. 1). In substance, condition monitoring is a classification problem [12]. The principal function of the condition monitoring is to check the operating condition of the system. It is made up of two parts which are detection and the diagnosis. The phase of detection makes it possible to determine the state of the system as being normal or abnormal. The phase of diagnosis consists in identifying the failing components and to find the causes starting from a whole of symptoms observed [7, 10, 12]. An industrial system is described by a vector of numeric or nominal features. Some of these features may be irrelevant or redundant. Avoiding irrelevant or redundant features is important because they may have a negative effect on the accuracy of the classifier [7,10]. In addition, by using fewer features we may reduce the cost of acquiring the data and improve the comprehensibility of the classification model (fig. 2). Fig. 1 Rotary Cement kiln Feature extraction and subset selection are some frequently used techniques in data pre-processing. Feature extraction is a process that extracts a set of new features from the original features through some functional mapping [15]. Subset selection is different from feature extraction in that no new features will be 8