Data visualization for data science in real estate market

Chua, Min Xuan (2026) Data visualization for data science in real estate market. Final Year Project, UTAR.

PDF Download (4Mb)

Abstract

Housing is one of the basic human needs, and real estate has always been in high demand. It is crucial for stakeholders to correctly estimate the value of a property so that they are able to make informed decisions. This research focuses on performing housing price prediction using machine learning and constructing a dashboard integrated with an automated ML pipeline. This research is based on a dataset consisting of Singapore's HDB public housing prices. The selected models for this research include Linear Regression, Decision Tree, Support Vector Machine, Gradient Boosting Machine, Random Forest, and Extreme Gradient Boosting. The necessary data preprocessing and feature engineering are performed, with customized column transformers applied to fulfil the need to handle high-cardinality columns. After the baseline model comparison, XGBoost, GBM, and Random Forest turned out to be the three best performers. These three models, along with the baseline model, are selected as candidates for hyperparameter tuning to select the final model, which will be the scope of Project 2. Meanwhile, SVM turned out to be the worst performer among the selected models. In the experiment, the DT showed significant overfitting. After hyperparameter tuning using Randomized Search and completing the comparison, XGBoost with its optimal parameters was selected as the final model. A dashboard was later constructed, supporting functionalities including data visualization, user CSV uploads, automated model training, and single or batch housing price prediction. This project contributed to furthering the research from theoretical results to an actual developed system for housing price prediction, with production efficiency considered, helping end users to make informed decisions.

Actions (login required)