An intelligent self-configuring recommender system as a service.

Abstract

In the rapidly evolving digital landscape, users are often overwhelmed by the multitude of listing services, ranging from music platforms to product recommenders and social media content suggestions, leading to a challenge in finding items that align with their individual preferences. To address this complexity, the development and implementation of Recommender Systems has become increasingly valuable. These systems efficiently sift through large volumes of data to match items with user preferences, thereby enhancing the choices available to users.

The focus of this work is on the development of an advanced Application Programming Interface (API) for Recommender Systems. This Application Programming Interface is uniquely designed to be universally accessible and easy to deploy. Serving as the backbone for various Web Services, the Application Programming Interface utilizes the robust Representational State Transfer architecture. It is crafted with an emphasis on modularity, promoting adaptability and flexibility. The Application Programming Interface processes user data and queries to deliver customized recommendations swiftly.

Performance evaluations have demonstrated the commendable accuracy of the Application Programming Interface. It exhibits outstanding performance particularly with smaller datasets, showcasing rapid data processing and algorithm execution times. The Application Programming Interface has shown exceptional efficiency and resilience under specific testing conditions, including cloud environments, and this is particularly evident in scenarios involving extensive datasets of up to 16,000 items. The Application Programming Interface is more than a mere tool; it represents a pathway towards personalized digital experiences, excelling in Create, Read, Update, and Delete operations and customized recommendations.

The user evaluation phase included a diverse group of participants, ranging from novice to experienced developers. Over half of these participants had substantial experience in software development, and a significant proportion had previously worked with coding recommender systems. Given their varied knowledge of recommender libraries, most feedback commended the effectiveness of the Application Programming Interface. 81% of users valued the recommendations provided, and many expressed confidence in its filtering techniques. The standout feature of this work is the versatility of the Recommender System Application Programming Interface.

Despite the positive feedback, users suggested improvements in areas such as documentation, data security, and features. These insights are valuable for future refinements of the Application Programming Interface and the user experience. The enthusiastic engagement and feedback from participants underscore the potential of the Application Programming Interface to enhance applications that require a recommendation system, particularly for developers who may not be as familiar with the theoretical aspects. The solid research foundation and the dedication of the participants highlight the potential for broader adoption of the Application Programming Interface by developers.