PropInsight: A Machine Learning and Visual Analytics Framework for Real Estate Price Prediction

Abstract

Estimating the prices of a homes is a very hard problem because there are so many different factors that an effect it. These factors include the home's location, size, furnishings, floor level, and a luxury grade. Standard valuation techniques that rely on a manual expert evaluation and subjective market comparison often produce inconsistent results, especially in a rapidly changing urban markets. This paper presents the PropInsight, a comprehensive machine learning pipeline and an interactive visual analytics dashboard designed to an automate and enhance real estate valuation for the Gurugram residential market. The system that scrapes property listings from the 99acres goes through a long preprocessing pipeline that gets rid of the duplicates, logically fills in a missing value, handles outliers, encodes features, and extracts useful engineered features. Then, the five regression models are trained and the tested: the Linear Regression, Random Forest, XGBoost, Gradient Boosting, and AdaBoost. The R2 score and the Mean Absolute Error (MAE) are used to measure performance. The experimental results show that the XGBoost has the best predictive accuracy (R2 = 0.908) followed by the linear baseline, which has a very different performance. The trained pipeline is stored in a Streamlit web app that can predict prices in real time and do exploratory analytics at the sector level. The app is called pipeline.pkl. This study demonstrates that the systematic integration of data engineering with comparative model evaluation and user-facing visualization constitutes a pipeline for a practically deployable, interpretable decision-support tool for stakeholders in the real estate sector.