Mammogram images are used to identify the breast cancer cells. The modern lifestyle changes and conventional medical diagnosis delays the process of identifying cancer cells at the early stages. Clinical practices and diagnostic techniques need further enhancements in processing cancer images. The number of features extracted for classification and the type of feature utilized for classification determine the accurate location of accurate cancer cells. Basically feature extraction inputs will be given for classification algorithms to detect the cancer cells, where a limited set of feature extraction parameters have been used to do the prediction. We propose a New Feature extraction (NFe) Algorithm with the major set of texture and geometrical parameters to improve the accuracy levels in predicting cancer images. The New Feature extraction (NFe) Algorithm is compared with existing methods and it is observed that there is a 2% improvement of accuracy for the proposed method. In order, to improve the accuracy of the classifier ensemble learning method is also employed. Our proposed mechanism is not only tested on objectively but also tested on subjectively with the view of expert medical physicians on real-time medical images. The tested results achieved 97% of closeness with subjective and objective testing. © 2019, Institute of Advanced Scientific Research, Inc. All rights reserved.