Resumen

COLMENARES-GUILLEN, Luis Enrique; CARRILLO-RUIZ, Maya; MORALES-MURILLO, Victor Giovanni  y  LOPEZ Y LOPEZ, José Gustavo. Validation of a classification algorithm for identifying pharmacological interactions. Ing. invest. y tecnol. [online]. 2019, vol.20, n.2. ISSN 2594-0732.  https://doi.org/10.22201/fi.25940732e.2019.20n2.014.

The pharmacological interaction is the modification of the effect of one drug by the action of another. In Mexico, the number of undesired drug reactions caused by pharmacological interactions has increasing year in year out. In this article an analysis of medical dictionaries such as iDoctus México, PLM México and Vademécum was carried out. With this analysis, a corpus with 540 pharmacological interactions that can occur in the most frequent medications at the Hospital Universitario de Puebla was developed. A classification model was generated on the Weka platform using a Naïve Bayes algorithm, which predicts the possibility of a pharmacological interaction classified as mild, moderate or severe. The tests with the Naïve Bayes algorithm were performed using the cross-validation method with 10 folds. Subsequently, the validation tests were compared with the results obtained by the Random Forest algorithm, again using the cross validation method with 10 folds. The results of the Naïve Bayes algorithm are 79.1%, in precision; 75.7% in recall, and 74.8%, in F-measure. With regard to the Random Forest algorithm, the data is as follows: 51.7%, in precision: 51.7% in recall, and 50.7%, in F-measure. Finally, the results obtained would benefit pharmacovigilance to approximate predicting new pharmacological interactions, prior to marketing a drug.

Palabras llave : Pharmacological Interaction; Pharmacovigilance; Learning Machine; Classifier; Validation.

        · resumen en Español     · texto en Español     · Español ( pdf )