Predicting structure use with machine learning algorithm: Model validation approach for DAP data

Abstract

Machine learning techniques have been successfully applied in many fields, including urban planning. The focus of this article is to develop a machine learning model to automatically predict the use of structures. Automatic predictions can help mitigate the heavy load on urban planners in the early stages of decision-making and provide a quick preview of the scenario. In this study, building data from the Detail Area Plan of Dhaka were used. The number of floors and basements in a structure, the structure's age, the number of dwelling units and the structure type were the independent variables for this research. Due to the dataset's inclusion of both numeric and string data, the Decision Tree (DT) classifier was used for prediction. Python routines were used for data cleaning, model development, and model evaluation. The Scikit-learn Python package, primarily used for ML implementation, was utilized to develop the model. The model had an accuracy rate of 91% for predicting the use of institutional, education and research, mixed use, health facilities, under construction, and agriculture structures. Due to incomplete data, residential, restricted and special use, community facilities, miscellaneous, commercial, industrial, transportation and communication use of structures could not be reliably predicted. This model can aid in determining the use of a structure based on the characteristics of the structure (floor, basement, structure type, structure age, dwelling unit), based on historical data for that location. The model demonstrates the use of machine learning in urban planning.

JBIP, Vol. 16, 2023, pp. 19-39