<![CDATA[Purpose – The rapid growth of urbanization has significantly increased demand for accessible, efficient boarding house ('kos') search platforms. However, most existing applications rely on static, manual filtering, which often leads to information overload and fails to capture implicit user preferences. To address this, this study designs and implements 'NK Kost,' a mobile application providing customized boarding house recommendations. Design/methods/approach – grates context-based search functionalities, including geolocation, radius, budget, room type, and amenities-with a server-side recommendation engine. To generate personalized results, the system evaluates three Collaborative Filtering (CF) techniques: User-Based CF, Item-Based CF, and a proposed Hybrid CF-Entropy model that combines user- and item-based predictions with entropy-inspired adaptive weighting, designed to address data sparsity and rating variability in localized boarding house contexts. Findings – Experimental results indicate that the proposed model achieves competitive, though not definitively superior, performance. User-Based CF achieved the lowest prediction error (RMSE = 0.90, MAE = 0.63), while Hybrid CF-Entropy demonstrated comparable ranking effectiveness (Precision@5 = 0.13, Recall@5 = 0.32) without outperforming the baselines across all metrics. A usability evaluation involving 35 participants yielded a System Usability Scale (SUS) score of 71.2, indicating acceptable perceived usability. Research implications/limitations – while the algorithmic adaptability of the Hybrid CF-Entropy model requires further validation on larger datasets. Originality/value – This study introduces a localized boarding house recommendation application using a Hybrid CF-Entropy model with adaptive weighting.]]>
