Wasila RehmanFollow

Date of Submission

Fall 2023

Supervisor

Dr. Tariq Mahmood, Professor and Program Coordinator MS(CS) and MS(DS) Programs, School of Mathematics and Computer Science (SMCS)

Committee Member 1

Dr. Tariq Mahmood, Program Coordinator MS (CS) and MS (DS), Institute of Business Administration (IBA), Karachi

Committee Member 2

Dr. Abdul Samad, Examiner – I, Institute of Business Administration (IBA), Karachi

Committee Member 3

Dr. Tahir Syed, Examiner – II, Institute of Business Administration (IBA), Karachi

Degree

Master of Science in Data Science

Department

Department of Computer Science

Faculty/ School

School of Mathematics and Computer Science (SMCS)

Keywords

Real Estate Appraisal, Street-View Images, Spatio-Temporal Housing Price Prediction, Deep Learning, Machine Learning

Abstract

Real estate appraisal is the technique to predict prices of a large number of properties that are similar together and is impossible to calculate without big geographical data. Appraisals are influenced by a multitude of factors, including the physical conditions of neighborhoods. When purchasing a property, abstract factors such as greenery, traffic, and commercialization tend to influence buyers’ purchase decision. Understanding how these physical conditions change over time and their impact on property prices is crucial for various stakeholders, such as homeowners, investors, and urban planners. Most researches have only focused on spatial features of an area and its impact on sale prices in a single year. The remaining researches have predicted property sale prices using temporal models. However, very little research has been done on how image data of a neighborhood varies over the years and its correlation with property prices. This study presents three ideas to overcome these issues: (1) designing a novel spatio-temporal data structure, the Spatial Temporal Abstract Detections (STAD), to quantify the influence of changes in the facilities on property prices, (2) designing a new framework to extract the most important features from street-view images to characterize property prices of each block in a neighborhood, and (3) building a novel YOLO-ML architecture to process Normalized STAD Scores (NSTADS) to predict prices. The study employs a comprehensive image dataset extracted from Mapillary, comprising physical attributes of neighborhoods, such as infrastructure, amenities, and environmental factors. The model’s performance is evaluated and compared with the baseline model to predict property prices.

Document Type

Restricted Access

Submission Type

Thesis

Download

Share

COinS