Descubrimiento de conocimiento en historias clínicas mediante minería de texto

The clinical institutions generate a large amount of unstructured data both in the registration of procedures in free text by medical staff, and by the images and videos generated by diagnostic aids. This paper proposes a process of knowledge discovery in the unstructured text of the medical records of the trauma area of the San Vicente Foundation Hospital through text mining. Text preparation techniques were applied such as elimination of non-relevant words, substitution of terms, elimination of accents and derivation of words. Regarding mining processes, supervised and unsupervised learning techniques were applied such as decision trees, logistic regression, nearest k-neighbors, hierarchical clustering and association rules. The result obtained is the conformation of a model of the most relevant words in the clinical records of the Hospital in the area of traumatology.