Machine learning techniques to identify patterns in gynecologic information

Medical records constitute an important source knowledge. Millions of data records can be processed looking for patterns using artificial intelligence and machine learning. Thus, the present research aims to identify patterns in gynecologic data. The dataset used includes 1251 records related to women's diseases, it contains aspects such as age, sicknesses, the contraceptive method used, and pathologic history, among others. The methodology applied in this work allowed the management of key aspects such as data understanding, preprocessing, modeling, and evaluation. Three unsupervised algorithms have been applied: k-means, DBSCAN, and Hierarchical Clustering. Silhouette metric has been used to evaluate the quality of each cluster. Results show that the best silhouette value was 0.73 and 9 clusters, obtained with DBSCAN. The outcomes obtained constitute an important contribution to identifying the most common genital infectious diseases that influence the identification of pattern in each cluster.