The Peshawar Property Puzzle: Predicting Prices in Hayatabad with Machine Learning

Abstract

The prediction of residential property prices has become an essential component of modern urban planning and real estate investment. This research focuses on predicting house prices in the Hayatabad region of Peshawar using a machine learning-based approach. A dataset of 1,300 housing records was collected and preprocessed to include structural, locational, and facility-related attributes. The target variable (house price) was categorized into three levels—low, medium, and high—to formulate a classification problem. Multiple algorithms were implemented and evaluated, including Logistic Regression (LR), Support Vector Machine (SVM), eXtreme Gradient Boost (XGBoost), and a proposed Multilayer Perceptron (MLP). A five-fold cross-validation strategy was adopted to ensure the reliability of the results. Among the tested models, the proposed MLP achieved the highest overall accuracy, demonstrating superior ability to learn nonlinear relationships between housing features and price categories. The findings indicate that ensemble-based models can provide more robust and precise predictions compared to linear approaches. This study contributes to the development of data-driven methods that support decision-making for real estate professionals, investors, and policymakers in regional housing markets