Breast cancer is the utmost dangerous deadly disease amongst women in the world. This disease detection in the early times is to be done by mammograms to increase the survival rate. Recent advancements in deep learning techniques are increasing the accuracy in discerning the disease in early stages. The objective of this article is mammogram classification by deep learning techniques, which contain several phases. The preprocessing phase is used to improve the contrast and to diminish noise in mammograms using Contrast Limited Adaptive Histogram Equalization (CLAHE), Gaussian filtering and median filtering. From the pre-processed images, features are extracted based on Advanced Gray-Level Co-occurrence Matrix (AGLCM). All these experiments are done on mammogram images that are available publicly. The classification is done using eXtreme Gradient Boosting (XGBoost) algorithm for classification. The performance is evaluated and obtained an accuracy of 96.8%, respectively. These results show that the deep learning-based mammogram image classification is very much helpful for radiologists.