Intelligent Fraudulent Property Listing Classification using Ensemble Learning Techniques

Abstract

The rapid digital transformation of the real estate sector has significantly increased the reliance on online property listing platforms, making the authenticity of listings a critical challenge. Fraudulent property advertisements not only lead to financial losses but also undermine user trust in these platforms. Traditionally, verification processes relied on manual moderation or simple rule-based techniques such as keyword filtering, price validation, and duplicate detection. While these methods offer basic screening, they are inefficient, time-consuming, and incapable of capturing complex fraud patterns in large-scale datasets. To overcome these limitations, this study proposes an intelligent hybrid ExtraNet framework for automated classification of real estate listings. The model integrates the Extra Trees Classifier (ETC) and customized Neural Networks (NN) to enhance detection accuracy. The system processes structured tabular data comprising essential property attributes, including price, location, area size, number of rooms, amenities, and textual descriptions. A robust preprocessing pipeline is employed to manage missing values, encode categorical variables, and normalize numerical features for improved model efficiency. For performance benchmarking, two baseline models Logistic Regression (LR) and K-Nearest Neighbors (KNN) are implemented and evaluated. Additionally, the system is deployed as an interactive desktop application using a Graphical User Interface (GUI) built with Tkinter and supported by a My Structured Query Language (MySQL) database for efficient data management. Experimental results demonstrate that the proposed hybrid ExtraNet model achieves a classification accuracy of 99.61%, significantly outperforming traditional models in identifying fraudulent listings and enhancing platform reliability.