Authors: Gerard Allali*, Manchester Metropolitan University
Topics: Urban Geography, Spatial Analysis & Modeling, Quantitative Methods
Keywords: Urban growth, Machine Learning, Cellular Automata, LightGBM, Support Vector Machine, Abidjan
Session Type: Virtual Poster
Start / End Time: 8:00 AM / 9:15 AM
Room: Virtual 26
Presentation File: Download
Urban growth in Sub-Saharan Africa is progressing rapidly since the 1960s. Abidjan city is one of the biggest metropolises in francophone region in West Africa and is following the same fast-growing development. To get a deeper knowledge of the city expansion and predict its future pattern, this case study adopts the remote sensing data, Machine Learning tools and Cellular Automata. The research methodology consists of firstly classifying LANDSAT images using LightGBM algorithm for the years 1987, 2000 and 2014, secondly evaluating the influence of the proximity to the road and the industrial centres, the slope of the terrain and the population density by applying the Geographical Weighted Logistic Regression and thirdly predicting the urban fabric of the city from 2000 to 2014 using the Cellular Automata with a transition rules based on the Machine Learning technique: Support Vector Machine. The research findings have demonstrated the good performance of the Gradient Boosting Machine algorithm LightGBM in the supervised classification of remote sensing data; proximity to roads and industrial centres have a positive contribution to urban growth while the slope and the population density are not positively linked to the city spreading; the assessment of the CA model prediction revealed an Area Under the Curve of 67.45% which is relatively accurate. The machine learning techniques are fit for purpose in the image classification and decision rules in the CA, but the researcher need to have a deeper knowledge of the input data in order to understand the output values from ML algorithms.