Enhancing Housing Price Prediction Accuracy Using Decision Tree
Regression with Multivariate Real Estate Attributes
Dublin Core
Title
Enhancing Housing Price Prediction Accuracy Using Decision Tree
Regression with Multivariate Real Estate Attributes
Regression with Multivariate Real Estate Attributes
Subject
House Price Prediction, Machine Learning, Decision Tree Regression, One-Hot Encoding.
Description
The real estate sector functions as a critical barometer of a nation’s economic performance; however, its inherent volatility and intricate pricing
mechanisms often hinder precise valuation—particularly in developing urban markets. In the context of Indonesia, where the property industry
contributes substantially to national GDP, deriving fair and data-driven housing price estimates remains a persistent challenge. Traditional
appraisal methods, which rely predominantly on subjective human judgment, frequently fall short in reflecting market dynamics accurately. This
research seeks to construct an interpretable machine learning framework for predicting residential housing prices by employing a Decision Tree
Regression (DTR) model. The DTR method was chosen for its transparent and hierarchical structure, allowing for a clear understanding of how
individual property characteristics affect price outcomes. The study utilizes a public dataset from Kaggle containing key housing attributes,
including land area, building size, number of rooms, and location variables. The methodological steps encompass data preprocessing (cleaning
and encoding using One-Hot Encoding), data partitioning into training and testing sets with an 80:20 ratio, and model performance evaluation
using standard regression metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and the Coefficient of Determination (R²).
The model attained an R² value of 0.385, suggesting that the selected features explain approximately 38.5% of the variance in housing prices.
While this indicates moderate predictive capability, the DTR model offers valuable interpretive insights—particularly in identifying land area as
the most influential predictor of price. The findings highlight that interpretable machine learning approaches can serve as effective analytical
tools for property valuation in emerging markets, balancing predictive accuracy with transparency. Moreover, this study lays the groundwork for
the future development of ensemble and hybrid predictive models, as well as the integration of AI-based analytics into decision-support systems
for property valuation, investment forecasting, and urban development planning in Indonesia’s evolving real estate landscape.
mechanisms often hinder precise valuation—particularly in developing urban markets. In the context of Indonesia, where the property industry
contributes substantially to national GDP, deriving fair and data-driven housing price estimates remains a persistent challenge. Traditional
appraisal methods, which rely predominantly on subjective human judgment, frequently fall short in reflecting market dynamics accurately. This
research seeks to construct an interpretable machine learning framework for predicting residential housing prices by employing a Decision Tree
Regression (DTR) model. The DTR method was chosen for its transparent and hierarchical structure, allowing for a clear understanding of how
individual property characteristics affect price outcomes. The study utilizes a public dataset from Kaggle containing key housing attributes,
including land area, building size, number of rooms, and location variables. The methodological steps encompass data preprocessing (cleaning
and encoding using One-Hot Encoding), data partitioning into training and testing sets with an 80:20 ratio, and model performance evaluation
using standard regression metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and the Coefficient of Determination (R²).
The model attained an R² value of 0.385, suggesting that the selected features explain approximately 38.5% of the variance in housing prices.
While this indicates moderate predictive capability, the DTR model offers valuable interpretive insights—particularly in identifying land area as
the most influential predictor of price. The findings highlight that interpretable machine learning approaches can serve as effective analytical
tools for property valuation in emerging markets, balancing predictive accuracy with transparency. Moreover, this study lays the groundwork for
the future development of ensemble and hybrid predictive models, as well as the integration of AI-based analytics into decision-support systems
for property valuation, investment forecasting, and urban development planning in Indonesia’s evolving real estate landscape.
Creator
Ahmar Dwi Utomo1,*
, B Herawan Hayadi2
, Eko Priyanto3
, B Herawan Hayadi2
, Eko Priyanto3
Source
https://ijiis.org/index.php/IJIIS/article/view/226/151
Publisher
University of AMIKOM Purwokerto,
Date
desember 2024
Contributor
Fajar bagus W
Format
PDF
Language
English
Type
Text
Files
Collection
Citation
Ahmar Dwi Utomo1,*
, B Herawan Hayadi2
, Eko Priyanto3
, “Enhancing Housing Price Prediction Accuracy Using Decision Tree
Regression with Multivariate Real Estate Attributes,” Repository Horizon University Indonesia, accessed January 2, 2026, https://repository.horizon.ac.id/items/show/9719.
Regression with Multivariate Real Estate Attributes,” Repository Horizon University Indonesia, accessed January 2, 2026, https://repository.horizon.ac.id/items/show/9719.