A customizable recommender system for mashup platforms
Carsten Radeck, Klaus Meißner · 2017
When developing composite web applications, non-programmers have to be assisted by recommendations. In the light of ad-hoc mashup development, users' heterogeneous skills and approaches, as well as different focus of mashup platforms, a variety of recommendation scenarios emerge. These require recommender systems to be flexible and configurable. However, recommender systems of prevalent mashup platforms lack sufficient customizability to such context-specific needs. In this paper, we introduce our novel recommender system. Therein, the general tasks a recommender has to perform, i. e., identify when recommendations are necessary, calculate, present and integrate these, can be declared and configured based on recommendation strategies, according to a certain purpose. Further, we present an adequate architecture, novel models and the overall workflow of our approach. Thus, our recommender is highly customizable and extensible for domain-and platform-specific needs, allowing to cope with a variety of scenarios. To show feasibility and suitability, we implemented our concepts within the CRUISE mashup platform.