Entwicklung eines hybriden Empfehlungssystems zur Optimierung der Wohnungsvermittlung: Integration von K-Nearest-Neighbors und Geschäftslogik bei der WAG Wohnungsanlagen GmbH

Abstract

The Austrian rental housing market confronts unprecedented challenges: soaring prices intersect with increasingly diverse applicant profiles. WAG Wohnungsanlagen GmbH, Upper Austria’s largest housing corporation, processes hundreds of monthly applications manually — an approach reaching operational limits. This thesis presents the development of a hybrid recommendation system integrating K-Nearest-Neighbors (KNN) algorithms with domain-specific business logic. The architecture introduces a manual boost mechanism prioritizing explicit preferences for specific properties. Feature engineering transforms heterogeneous apartment and applicant data into normalized vectors; MinMaxScaler enables meaningful distance calculations. Empirical validation using authentic allocation data (May-June 2025, n=47) reveals substantial improvements. Pure KNN achieved 6.38% accuracy for top-10 recommendations; the hybrid approach reached 23.40% — a 267% increase. Among successful allocations, 72.73% originated from business logic integration, validating the importance of domain expertise. Mean Reciprocal Rank rose from 0.031 to 0.201. Implementation employs SQL for data preparation and Python for algorithmic processing. One-hot encoding transforms 28 settlement areas; context-based imputation handles missing values. The architecture scales from hundreds to thousands of applicants while maintaining transparency and regulatory compliance. Results demonstrate that combining human judgment with algorithmic efficiency surpasses either approach alone. Future work might explore temporal preference dynamics, fairness metrics for social housing, and multi-objective optimization balancing tenant needs with provider constraints. This research provides foundations for digital transformation in commercial housing while preserving social mission objectives.

Date of Award	2025
Original language	German (Austria)
Supervisor	Gabriel Kronberger (Supervisor)