Boosting API Recommendation With Implicit Feedback

Developers often need to use appropriate APIs program efficiently, but it is usually a difficult task identify the exact one they from vast of candidates. To ease burden, multitude API recommendation approaches have been proposed. However, most currently available recommenders do not support effective integration users' feedback into loop. In this paper, we propose framework, BRAID (Boosting RecommendAtion with Implicit FeeDback), which leverages learning-to-rank and active learning techniques boost performance. By exploiting information, train model re-rank results. addition, speed up process learning. Existing query-based can be plugged BRAID. We select three state-of-the-art as baselines demonstrate performance enhancement measured by Hit@k (Top-k), MAP, MRR. Empirical experiments show that, acceptable overheads, improves steadily substantially increasing percentage data, comparing baselines.