Browsing by Author "Zabala Cárdenas, Anthony Roger"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Pronóstico de precipitaciones pluviométricas usando el modelo univariante Prophet para una toma de decisiones informada.
    (Universidad Técnica de Ambato. Facultad de Ingeniería Civil y Mecánica, Carrera de Ingeniería Civil., 2025-02) Zabala Cárdenas, Anthony Roger; Viscaíno Cuzco, Mayra Alexandra
    Meteorological information is essential to understand the climatic state of a region and becomes a key tool for decision-making in various sectors of society. During periods of water scarcity, it is essential to implement water management systems that mitigate the impact of shortages. In this context, the main objective of this research was to forecast rainfall in the Ambato canton using the Prophet univariate model for informed decision-making. This work was structured in three phases with the purpose of implementing the Prophet model to make daily rainfall forecasts in the Ambato canton. In the first phase, a statistical analysis of the historical series was carried out, calculating measures such as average, quartiles and standard deviation. The second involved deploying the Prophet model at 19 weather stations with records between 2013 and 2024, splitting the data into proportions of 80 and 20 percent, 85 and 15 percent, and 90 and 10 percent, using the first set to train the model and the second to evaluate the results. In addition, the hyperparameters changepoint prior scale and seasonality prior scale were adjusted, with ranges between 0.05-10 and 10-100, respectively, using grid search to determine the optimal configuration that minimized MAE and RMSE errors, comparing it with the default values of the model. Finally, the third consisted of creating isohyet maps that illustrate the spatial distribution of precipitation. The results showed that, at most stations, the Prophet model fit the data split of 90 percent for training and 10 percent for evaluation more effectively, obtaining average MAE and RMSE values of 2.13 mm and 3.41 mm, respectively. This allowed us to conclude that the Prophet model demonstrated a good performance in fitting the analyzed time series. However, it had difficulties in predicting maximum precipitation values, which is a key aspect to consider in future applications of the model.