Hybrid Feature Extraction and Selection Using Bayesian Classifier Myneni Madhubala and Dr. M. Seetha Associate Professor, Dept. of CSE, Anjamma Agireddy Engineering College for Women, Bandlaguda, Hyderabad 500 005, India. e-mail: baladandamudi@gmail.com Professor, Dept. of CSE, G. Narayanamma Institute of Technology and Science for Women, Shaikpet, Hyderabad 500 008, India. e-mail: smaddala2000@yahoo.com Abstract. This paper presents a image classification using Bayesian approach for hybrid feature selection and classification of image data without expert interaction. Our method defines a new hybrid feature subset is able to classify the data without expert knowledge. The proposed method is learned by using the probabilistic bayes net classifier algorithm with less semantic details. The experimental results will show the merits of the proposed methodology in the classifying image data. Keywords. Hybrid Features, Image Classification, Image Mining, Bayesian Network. 1. Introduction With image mining, consider the four broad problem areas associated with data mining: finding associ- ations, classification, sequential patterns and time series patterns. Image mining has two challenging issues like creating image feature attribute file and discrediting the data into higher levels. For creat- ing image attribute file feature selection is one of the important tasks. Image is represented in intensity values at pixel level of image dimensions. Working at image pixel level is complex issue. The next level representation of image is feature level. We had bag of features available to represent an image. The basic primitive features are based on colour, texture and shape. Any one of these feature is not sufficient to get high performance result. Now in this paper hybrid features are introduced with the combination of color, texture and shape. Hybrid feature subset is used on optic nerve images [13]. In this work three feature extraction algorithms are used. The first color feature extraction algorithms are based on principal component transformation to convert RGB image to HSV image. Secondly for texture feature extraction, the pyramid-structured wavelet transform is used. Thirdly edge feature extraction has done with canny edge extraction. The next task is extracted features are represented in vector form to generate image attribute file. After applying image mining classifi- cation algorithm the performance has been evaluated. The Bayesian network algorithm has been used for feature selection [4]. 2. Image Data The image data used to do the classification has 3 categories in the experiment: Tankers, Missiles, Flowers and animals, each of them containing 100 positive images and 100 negative images (randomly collected from other categories). All images are in JPEG format with any size. In the preprocessing stage all the images are resize to 256 × 256. Then transfer each image into a feature vector and write them into a single data file in the specified format. The image feature vectors are generated by using familiar feature extraction algorithms. Various encoding strate- gies such as normalization (e.g., adjusting the value ranging from 0 to 1) or generalization (e.g., transform- ing the value to high, medium, or low) can be applied when generating the desired features. 3. Methodology The methodology of image mining includes various steps viz. feature extraction, feature selection using learning, transformation, training the classification model and test with new data. The methodology used for classification has shown in Figure 1. 449