Simulations, visualisations and analytics
There are various smaller-scale projects I did throughout my university life. These include Arena models, which made me look at the world of optimisation differently, a data visualisation project built in Tableau, and a Kaggle competition done as a final team project in Singapore.
Arena models
Despite having already taken an Operations Management course during my bachelor's, it was only when I completed the Simulation Theory and Optimisation master's course that I discovered my love for optimising processes. It introduced me to discrete-event simulation and sparked a sincere interest in trying out different variables and paths while designing models to optimise production, queues and waiting times. I used several of Arena's integrated tools — the Process Analyzer to compare scenarios, OptQuest to optimise variables such as the number of resources needed and the schedules used, and the Input and Output Analyzers to fit and study the data. My experience with the Arena simulation software mainly comes from the pre-prepared projects in the course book, along with the final team project. It showed me the practical side of simulation and optimisation, and how important data, control and queue management are for an organisation.
A sketch of one of the Arena projects.
Data visualisation of Agora's booking system
When completing the Data Visualisation course during my Master of Information Management at KU Leuven, my team and I came up with an original idea for the final project to visualise seat booking data from one of the most well-known libraries in Leuven: Agora.
Main Problem: During peak study periods, especially before exams, it is extremely difficult for students to find a study seat. The visualisation was made to inform students about booking trends and help them make better-informed seat booking decisions. I was responsible for extracting, processing and visualising data related to booking cancellations. Since seats may become available either because of user cancellations or no-shows, my visualisation helps by identifying the exact dates, hours and minutes when most cancellations are likely to occur, thereby increasing students' chances of securing a seat. The visualisation was created in the form of interactive histograms.
Preview of cancellation behaviour visualisation
Kaggle final project at SMU
The final assignment of one of the two subjects during my summer programme at SMU was a Kaggle competition done in teams. The goal was to forecast daily sales for thousands of grocery items across the stores of a large Ecuadorian retailer. As a team of five, we built an AI (LightGBM) model driven by heavy feature engineering — lagged sales, moving averages, holiday and oil-price signals — and validated it with time-series cross-validation. The process required both a deep analysis of the raw data and the development of a machine learning model. It scored an RMSLE (the error metric used in the competition) of 0.426 and placed 76th of 733 in the world. The project taught me the importance of the right data: the final model did not use every piece of data available, but instead studied the relationships and hand-picked the greatest contributors, rather than overloading it.
Feature importance of the model.