Towards Reliable IoT and Web API Mashups: A History-Aware Linked Data Composition Approach

Abstract

The growing number of heterogeneous Web and IoT APIs poses a fundamental challenge to the reliability and reusability of modern applications, as existing API composition frameworks are hindered by semantic fragmentation and a neglect of historical usage patterns. To address this problem, this research paper introduces a history-aware semantic framework that enhances the long-term robustness of service compositions by embedding APIs into a unified knowledge graph via a dedicated ontology that harmonizes heterogeneous data and services through semantic abstraction and domain-level alignment. Reliability of composition is further enhanced by incorporating historical composition data capturing co-occurrence and popularity. A genetic algorithm dynamically optimizes the weights of these metrics to select the Top-K most reliable mashup compositions. Extensive evaluation on the large-scale ProgrammableWeb benchmark (11,492 APIs and 7,336 mashups) demonstrates a statistically significant improvement of 26% in mashup reliability (p < 0.01) compared to state-of-the-art baselines. While this work focuses on Intelligent Media mashups as a representative application domain, the proposed framework is applicable to general Web and IoT API composition. The integration of semantic abstraction, historical analysis, and optimization provides a robust solution to generate resilient, adaptive, and highly reliable service mashups, offering a useful approach to support the long-term adaptability and reuse of complex service ecosystems.