border border

Optimisation Engine for Site Search

Team members

Zach Ang Zi Jie (ESD), Ignasia Hanny (ESD), Jason Andika Lie (ESD), Teo Wei Jie Zachary (ISTD), Tan Yu Xiang (ISTD), Amanda Kosim (ISTD)


Yue Mu, Oka Kurniawan

Writing Instructors:

Nurul Wahidah Binte Mohd Tambee

Teaching Assistant:

Siddharth Kumar

gradsingapore logo ura 2019


Urban Redevelopment Authority (URA) is responsible for identifying suitable land plots for siting of new amenities. This process is known as Site Search. The current process of site search relies a lot on the planner’s familiarity. Local planners can spend a dedicated 1 to 2 weeks of work to identify suitable potential sites, using many different applications and data.


Finding new sites for facilities is a multi-faceted problem that involves the matching of time, space and various planning considerations. Planners have to understand and analyse the:

    ⦿ Demographics of an area for the particular timeframe

    ⦿ The capacity of a facility

    ⦿ Catchment of the facility

    ⦿ Added considerations specific to a site


The challenge increases exponentially in complexity when planners look for sites with a combination of requirements. Trade-offs then have to be weighed against each other, adding to the challenge.

Hence, there is value to exploring an evidence-based decision making support system, to help planners suggest options for site search.



Problem Statement

The aim of our project is to design a data-driven recommendation engine for site search.

This engine should optimise the quantitative aspects of the search, recommending optimal potential sites to the planners for their qualitative analysis.




Optimisation Models


Our solution includes three different objective functions which are inspired by different facility allocation models. These are solved using the exact method, with the aid of a powerful commercial solver, Gurobi.

optimisation models


The Multi-Objective Optimisation is solved using a genetic algorithm.
Planners can select any combinations of the 3 models which is then passed to the script to perform optimisation.

NSGA-II Genetic Algorithm

The NSGA-II Genetic Algorithm iteratively improves the pareto front by performing a ranking algorithm.

Pareto front is the set of non-dominated solutions. Each solution is better than or equal to all other solutions within the set for at least 1 objective.

The best pareto front is then shown on the web application as a combination of solutions.



student Zach Ang Zi Jie Engineering Systems and Design
student Ignasia Hanny Engineering Systems and Design
student Jason Andika Lie Engineering Systems and Design
student Teo Wei Jie Zachary Information Systems Technology and Design
student Tan Yu Xiang Information Systems Technology and Design
student Amanda Kosim Information Systems Technology and Design
border border