MDE-Based Approach for Accelerating the Development of Recommender Systems in Smart Tourism

Abstract

Recommender systems (RSs) have become fundamental computational tools deployed across diverse domains, including e-commerce, tourism, and streaming platforms to facilitate personalized content delivery through the analysis of user preferences, behavioral patterns, and interaction data. With the use of Machine Learning (ML) techniques, these RSs can be implemented either through established frameworks such as NReco and Apache Mahout, or through custom development. Despite the availability of multiple implementation approaches, RS development remains computationally complex, necessitating substantial expertise in both software engineering principles and artificial intelligence methodologies. To address these technical challenges, this paper presents a Domain-Specific Language (DSL) framework that streamlines and accelerates RS development for smart tourism applications through a Model-Driven Engineering (MDE) methodology. The proposed framework minimizes the requirement for extensive programming or ML expertise and enables developers to efficiently generate customized RS implementations while maintaining system quality. The DSL framework automates the generation of code that leverages the services provided by the Apache Mahout framework for implementing recommendation algorithms, abstracting the underlying complexity and enabling developers to focus on high-level system design rather than the technical details of algorithm implementation. In addition, the framework integrates a code generator and a modeling tool, allowing developers to design and implement recommendation systems with reduced complexity and enhanced productivity. Unlike generic or low-code platforms, this approach combines domain-specific abstractions with automation tools to address the unique challenges of the smart tourism sector, ensuring adaptability, efficiency, and high-quality results.