Análisis de intención de compra en canales digitales mediante técnicas de explainable artificial intelligence (XAI)
Loading...
Download
Official URL
Full text at PDC
Publication date
2024
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
En el contexto del comercio electrónico en expansión, entender y predecir la intención de compra es crucial para personalizar campañas y aumentar las conversiones. Sin embargo, la opacidad de los modelos “caja negra” dificulta el análisis detallado y las decisiones informadas. Este estudio utiliza técnicas de Explainable Artificial Intelligence (XAI) para abordar estos desafíos en el marketing digital, enfocándose en identificar los factores que influyen en las intenciones de compra en línea. Los objetivos incluyen construir modelos predictivos mediante validación cruzada y una metodología basada en KDD, aplicando técnicas globales de XAI para identificar variables clave y métodos locales para explicar casos específicos. Se presenta un modelo XGBoost con un umbral de 0.4, alcanzando un ROC AUC de 0.9268 y una precisión del 0.9007 en los datos de prueba. Los resultados subrayan la importancia de métricas como PageValues y ExitRates de Google Analytics, variables temporales como el mes y la actividad del usuario en páginas de productos relacionados, como elementos críticos para entender las decisiones de los usuarios. Basado en estos hallazgos, se recomienda mejorar el contenido y la funcionalidad de las páginas web, optimizar páginas clave para retener usuarios, intensificar campañas durante períodos de alta conversión, y mejorar la interfaz del cliente para una experiencia más satisfactoria.
In the context of expanding e-commerce, understanding and predicting purchase intent is crucial for customizing campaigns and increasing conversions. However, the opacity of “black box” models complicates detailed analysis and informed decision-making. This study employs Explainable Artificial Intelligence (XAI) techniques to address these challenges in digital marketing, focusing on identifying factors influencing online purchase intentions. Objectives include building predictive models through cross-validation and a KDD-based methodology, applying global XAI techniques to identify key variables, and using local XAI methods to explain specific cases. An XGBoost model with a threshold of 0.4 is presented, achieving a ROC AUC of 0.9268 and a precision of 0.9007 on test data. Results underscore the importance of metrics like Google Analytics’ PageValues and ExitRates, temporal variables such as month, and user activity on related product pages as critical elements in understanding user decisions. Based on these findings, recommendations include enhancing website content and functionality, optimizing key pages for user retention, intensifying campaigns during peak conversion periods, and improving the client interface for a more satisfactory experience.
In the context of expanding e-commerce, understanding and predicting purchase intent is crucial for customizing campaigns and increasing conversions. However, the opacity of “black box” models complicates detailed analysis and informed decision-making. This study employs Explainable Artificial Intelligence (XAI) techniques to address these challenges in digital marketing, focusing on identifying factors influencing online purchase intentions. Objectives include building predictive models through cross-validation and a KDD-based methodology, applying global XAI techniques to identify key variables, and using local XAI methods to explain specific cases. An XGBoost model with a threshold of 0.4 is presented, achieving a ROC AUC of 0.9268 and a precision of 0.9007 on test data. Results underscore the importance of metrics like Google Analytics’ PageValues and ExitRates, temporal variables such as month, and user activity on related product pages as critical elements in understanding user decisions. Based on these findings, recommendations include enhancing website content and functionality, optimizing key pages for user retention, intensifying campaigns during peak conversion periods, and improving the client interface for a more satisfactory experience.