Breast Cancer is one of the second most common cancer among women and it is considered to be one of the major cause of cancer-related fatalities. This research incorporates the merits of an advanced Computer Aided Diagnosis (CAD) tool that has potentiality of automatic detection of mass and finding the abnormalities on lumps and ducts among mammogram images. This research presents a novel deep learning approach integrated Improved SqueezeNet with Adam optimizer to improve detection as well as classification accuracy. This proposed scheme follows a pipeline method for diagnosis of breast cancer. As a first step, Wavelet Transform with Contrast-Limited Adaptive Histogram Equalization (CLAHE) techniques are employed to eliminate the noise from the image and enhance the mass regions. Then, the Adaptive Thresholding and Contours techniques are used for selecting the region of interest from mammogram images, later an Improved SqueezeNet model is incorporated for training, Local Binary and Gray Level Co-occurrence Matrix are incorporated for extracting statistical features. Finally, the extracted statistical features undergo classification using various classifiers like Support Vector Machine, Random Forest, KNN and XGBoost Whereas, spatial features are classified using SqueezeNet classifier. Simulations of the method are applied to the Breast Tumor Mammography Dataset for Computer Vision dataset and the experimental results of SqueezeNet and Support Vector Machine exhibit superior performance when compared with other classifiers and the performance of the Improved SqueezeNet model shows better accuracy when compared with other models like ResNet50, Inception Net and VGG network models.