Ensayo en modelos predictivos de Data Mining para diabetes en etapa temprana

• Autor: Leiva Quispe, José Enrique
• Formato: tesis de grado
• Fecha de Publicación: 2023
• Institución: Universidad Nacional de Trujillo
• Repositorio: UNITRU-Tesis
• Lenguaje: español
• OAI Identifier: oai:dspace.unitru.edu.pe:20.500.14414/18558
• Enlace del recurso: https://hdl.handle.net/20.500.14414/18558
• Nivel de acceso: acceso abierto
• Materia: Riesgo de diabetes; Etapa prematura; Minería de datos; Regresión logística; Máquina de Soporte Vectorial; Métricas de evaluación

Diabetes has become such a common, but deadly, chronic health problem that it has _x000D_ increased dramatically in recent years. About 50% of all people with diabetes are not _x000D_ diagnosed due to its long-term asymptomatic phase, which is why detecting diabetes in an _x000D_ early phase is of vital importance. Science has advanced so much in the field of health that _x000D_ data mining classification techniques have been well accepted by the scientific community _x000D_ for the predictive model of disease risk. In the present investigation, a set of 520 data has _x000D_ been used, which information was collected through a direct survey of patients from the _x000D_ Sylhet Diabetes Hospital in Bangladesh. The respective analysis was carried out using _x000D_ classification algorithms such as Logistic Regression (classical statistical technique) and _x000D_ Support Vector Machine (machine learning technique). After adjusting the models and _x000D_ evaluating using metrics such as accuracy, sensitivity and AUC (in that order), it has been _x000D_ found that the Vector Support Machine model has a better fit and predictive power (0.98, _x000D_ 0.98, 0.99) compared to the model of logistic regression (0.92, 0.94, 0.97). Finally, as a _x000D_ suggestion, useful tips were raised to control risk factors