Browsing by Author "Perdomo Unaucho, Ana Gabriela"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Diseño de un modelo inteligente basado en aprendizaje automático para la predicción de enfermedades y mejora del proceso de atención en la clínica veterinaria SIRA PETS
    (Universidad Técnica de Cotopaxi (UTC). Ciencias de la Ingeniería y Aplicadas. Industrial, 2026-10-14) Chiluisa Gutiérrez, Bryan Alexander; Perdomo Unaucho, Ana Gabriela; Naranjo Robalino, José Ezequiel
    The study addressed the limitations of clinical information management at SIRA PETS Veterinary Clinic, which were characterized by scattered and non-standardized records that hindered access to patient history and the identification of recurrent diseases. Its objective was to design an intelligent machine learning-based model to predict canine diseases and improve the care process. A quantitative approach was applied through retrospective document review, observation, data cleaning, Pareto analysis, and supervised classification. The population comprised 500 canine clinical records; 396 records corresponding to the seven most frequent diseases were selected, representing 79.2% of the total. Twenty-five demographic and pre-diagnostic clinical variables were used, and five-fold stratified cross-validation was applied. The J48 decision tree, Random Forest, Naive Bayes, and support vector machine algorithms with a radial basis function kernel were compared using classification metrics. Random Forest achieved the best performance, with an accuracy of 0.9368 and a weighted F1-score of 0.9328. The functional test conducted with 72 additional cases generated predictions for the seven diseases and showed an initial agreement of 88.89% with veterinary judgment; this constituted an initial functional validation rather than an external clinical validation. As an improvement proposal, a standardized clinical record form and a dashboard developed in Streamlit were integrated to organize data entry, generate advisory predictions, and support veterinary assessment without replacing professional judgment. In addition, the proposal made it possible to estimate the percentage improvement in the care process by comparing the current process with the proposed process.