Comparativa entre modelos de árboles de clasificación y regresión para predecir la parasitosis intestinal en niños de 5 a 9 años en el Hospital Pedíatrico Alfonso Villagómez Román

Abstract

This research work compares two classification techniques: Classification Trees and Logistic Regression to predict intestinal parasitism in children from 5 to 9 years of age treated at the Alfonso Villagómez Román Pediatric Hospital in the city of Riobamba, in order to establish the best model whose associated factors are significant for the variable under study. The database for the analysis was provided by the Statistics Area of the health home, through the existing repositories, information was taken from the period 2019 -2021, for the information processing, those records that could present inconsistencies were purged, the validation for the new matrix was performed by comparing the clinical histories of the users, obtaining a new record with 2675 patients and 14 variables of interest. The models found were obtained using the SPSS software and its modeler SPSS MODELER, used to evaluate the predictive capacity through the goodness-of-fit measures: Error rate and Roc Curve; with this analysis, the classification tree model was the best for having a general precision of 65.7% and an error rate of 24.49%. The factors associated with the diagnosis are: Triage, Oxygen saturation, Axillary temperature, Height, Weight, Minimum heart rate and Type of health insurance. It should be noted that the difference in prediction efficiency for each technique was minimal, so it would be feasible to test other techniques, either by following the line of decision trees or regression models.