Breast cancer is one of the most common cancers among women worldwide, affecting around 10% of all women. Its fatality rate is extraordinarily high when compared to other cancer types. Histopathological analysis is one of the reliable methods for identifying breast cancer with confidence. But it's a highly specialized, time-consuming procedure that depends on the pathologist's experience. An automatic computerized histopathological image classification model can assist doctors and pathologists for early diagnosis. But most of the work is based on the non-interpretable black box model for feature selection tasks which is not very encouraging in the medical field. To introduce features interpretability in histopathological imaging, handcrafted features analysis is important first. In this paper, a novel hybrid computer-aided multi-class classification framework is proposed for predicting breast abnormality using histopathological images and handcrafted features. The proposed framework uses traditional handcrafted features (i.e. texture features, colour features, and fusion of texture and colour features) and a deep neural network (DNN) for breast cancer abnormality classification in order to facilitate the screening process for early diagnosis. The DNN classifier is trained using the features that were extracted using the handcrafted methods for two-class and multi-class classifications. The experimental results show that the use of handcrafted texture features had a better prediction performance in histopathological image classification as compared to the other handcrafted features for breast abnormality prediction. Using texture features with a DNN classifier, we achieved average values for accuracy, recall, precision and F1-score of 73%, 79%, 81% and 82% respectively for multi-class classification.