Modelo de predicción de parámetros de calidad de productos vínicos utilizando machine learning : Modelo de predicción de calidad de vino utilizando aprendizaje automático.

Abstract

The titling project seeks to determine whether it is possible to predict the quality of wine from its physical and chemical properties using an automatic learning model. Initially, bibliographic research was carried out using key terms related to the topic. Automatic learning models were collected, and a database was obtained for model training and testing. After comparing the models, it was decided to use neural networks and adjust their hyper parameters using the parameter grid process to improve their accuracy. Subsequently, database parameters were analyzed to identify the most influential characteristics in the model. Lime and Shape methods were used to explain the model and determine the most important physical and chemical properties for each possible quality value of the wine. Finally, the trained model was integrated into a view-control architecture system, with backend and frontend components. The backend was developed with Fastapi and frontend with JavaScript and Angular. This system will enable entrepreneurs to assess the quality of the wine produced and to obtain predictions about its quality.