This paper discusses automated breast cancer diagnosis based on histological images. The dataset consists of four different groups: normal tissue, benign carcinoma, in situ carcinoma and invasive carcinoma. We developed two algorithms to classify the images into these categories. Both include a preprocessing stage for noise elimination and cell segmentation, extraction of features and final diagnosis of the tissue along with malignity degree. The diagnosis is executed by classification using k-means, random forests and support vector machines. The best experiment resulted in an ACA of 0.475.