Urban Analytics
Philadelphia Census Block Group Gentrification Prediction Using Machine Learning
Utilized machine learning to forecast gentrification patterns in Philadelphia census block groups, addressing the socioeconomic impacts on urban development, housing affordability, and community dynamics. This project leveraged 2018–2019 data for training and validation to provide policymakers and planners with actionable insights for equitable growth and inclusive development.
Predict Housing Value Research Part 1: Seattle, WA
Developed a predictive model for housing prices in Seattle by analyzing variables such as property condition, structural characteristics, spatial location, and proximity to local amenities. The model provides insights into factors influencing housing prices, supporting stakeholders in the real estate market.
Predict Housing Value Research Part 2: Using Ordinary Least Squares (OLS) Regression to Predict Median House Values in Philadelphia
Conducted an OLS regression analysis to predict median house values in Philadelphia, examining key factors such as educational attainment, vacancy rates, single-family home proportions, and poverty levels. The study provides insights to inform housing policy, urban planning, and equitable development initiatives.
Predict Housing Value Research Part 3: Using Geographically Weighted Regression (GWR), Spatial Lag, and Spatial Error to Predict Median House Values in Philadelphia
Analyzed median house values in Philadelphia using OLS regression and advanced spatial econometric techniques, including geographically weighted regression (GWR) and spatial lag models. The study examined the impact of socio-economic factors such as educational attainment, vacancy rates, and poverty levels while addressing spatial dependencies to inform equitable and sustainable housing development strategies.
Predictive Policing
This project aims to predict robbery risk in Chicago, Illinois, by leveraging data from 311 service requests to identify spatial patterns and potential hotspots. The model incorporates features derived from data aggregated within each grid cell and calculates the distances from each cell to the five nearest data points to capture spatial relationships. Using this information, a predictive model is developed to assess robbery risk across the city.
Predicting Car Crashes Caused by Alcohol Using Logistic Regression
This project investigates the predictors of alcohol-impaired driving crashes in the United States, focusing on Philadelphia as a case study. Using logistic regression, the study examines various behavioral and socio-economic factors contributing to drunk driving incidents. Key predictors include driver age groups (teenagers and seniors), overturned vehicles, cell phone use, speeding, and aggressive driving, alongside socio-economic indicators like education levels and median household income. By analyzing these variables, the project aims to identify patterns associated with impaired driving and provide insights to inform safety interventions and policies.
Strategic Optimization of Police Patrol Locations Using Spatial-Temporal Analysis of Crime Severity Distribution
This study optimizes police patrol locations using spatial-temporal analysis of crime severity in Philadelphia. Crime data were assigned severity scores based on federal standards. Two models were used: the Maximum Coverage Location Problem (MCLP) to maximize coverage of high-severity crimes within a response distance, and the P-Median model to minimize the total weighted distance to all crimes. The findings reveal spatial-temporal variations in crime severity, influencing resource allocation throughout the day. The analysis highlights trade-offs between maximizing coverage and minimizing response times, providing insights for strategic deployment of police resources to enhance public safety.