Close
Research Title
An Integrated Deep Learning Model using Retail, Residential and Footfall Change data to Predict Retail Gentrification
- More about Byeonghwa
Academic Background
- Kyungpook National University : BA Geography (2008 ~ 2015)
- The University of Edinburgh : MSc Geographical Information Sciences (2017~2018)
- Korea Research Institute for Human Settlement : Assistant Research Fellow (2018~2019)
- University College London : Ph.D. student at the Department of Geography
- Publications
- ByeongHwa, Jeong, and Jung-Sup Um. "Comparative evaluation of South versus North Goseong County in terms of forest carbon offset potential." Spatial Information Research 24.1 (2016).
- ByeongHwa, Jeong, and JunWoo, Kim. “Application and Evaluation of Self-Organising Map-based Spatial Clustering for Regional Policy Recommendations.” Journal of the Korean Geographical Society 54.3 (2019):387-404.
- ByeongHwa, Jeong, and JunWoo, Kim. “Topic Modeling of the Transition of the Sense of Place in Urban Regeneration Area: A Case Study of Daegu Bangcheon Market. ” Journal of Daegu Gyeongbuk Studies 19.1 (2020):27-44
- Research Interests
Contemporary retail gentrification is widely considered to hinder urban sustainability and trigger social discord in cities in the Global North. The central motivation for this thesis is to predict the British high streets in which retail gentrification is likely to occur.
Retail gentrification is commonly observed to entail changes in retail composition, with new retail offers patronised by the young and middle class, as well as an increase in the volume of footfall. To estimate these changes, this study employs innovative and continuously updated datasets supplied by ESRC’s Consumer Data Research Centre (CDRC). These data make it possible to develop detailed representations of the preconditions to and processes of retail gentrification in terms of retail and residential change as well as footfall activity. The study period includes the COVID-19 pandemic, and a series of analyses are developed to assess any consequential and enduring effects upon retail gentrification.
We project changes in retail and high street structure and activities between the present day and 2040 using a deep learning Retail, Residential and Footfall (REF) model. This is comprised of Autoencoder, Long/Short-Term and Bayesian Neural Network components. Using its predictions, gentrified high streets in England are identified by investigating high streets that exhibit changes in composition and function. We also explore the geographical distribution of these locations as well as their noteworthy characteristics.
Taken together, this research develops a retail gentrification model using unique data and state-of-the-art deep learning techniques. It also explores the implications of the modelling effort for managing the phenomenon of retail gentrification through policy interventions.
