Un sistema web basado en grafología y machine learning para recomendación de parejas afines sentimentalmente : desarrollo del backend , modelo y Front End.

Abstract

The project develops an innovative web system that integrates a machine learning model to classify handwritten signatures into four personality types: bilious,nervous,lymphatic,and sanguine.Using the CRISP-DM methodology for model creation and Kanban for software development,the project addresses the lack of graphology professionals in Ecuador by providing a tool that allows personality evaluations and the recommendation of psychologically compatible individuals.The machine learning model was trained using the transfer learning technique with Inception V3,leveraging a pre-trained model to optimize resources.Model evaluations showed consistent improvements in accuracy and loss,and metrics indicated an acceptable level of classification,though with areas for improvement in the bilious and lymphatic classes.The implementation of Kanban facilitated agile and efficient project management,enhancing team productivity.The resulting web system allows users to upload images of their signatures and obtain a detailed personality analysis,as well as identify compatible individuals for long-lasting and satisfying relationships.The API deployment and model integration into the cloud infrastructure ensure the system's availability and scalability.