Jean-Luc MAHOROMEZA

Time series forecasting

Implementation of a long-term passenger traffic forecasting model Designed and developed a statistical and machine learning engine for time series forecasting of passenger traffic with a 12-month horizon. Built a multivariate, hierarchical model that ensures consistency across multiple aggregation levels (flights, terminals, airports). Applied advanced methods including state-space models, Bayesian hierarchical modeling, and online learning to continuously update forecasts as new data became available. Delivered a robust forecasting solution to anticipate passenger flows, improve operational planning, and support strategic decision-making. Keywords: time series forecasting, long-term prediction, hierarchical models, state-space models, online learning, passenger traffic, airports.

Probabilistic Adaptive and Multi-Horizon Time Series Forecasting. Laboratory of Probabilités, Statistique et Modélisation (LPSM)

• Building solutions using data. As part of the Cloud Enterprises solution team, I respond to customer issues using data. Loving mathematics applied to data, I explore this data, then I identify the law of data behaviour, then I make hypotheses to explain these hypotheses, and in the end, I test and deploy the solution to the business users. The solution in the form of a Machine Learning model is built using the resources and services offered by Google on Google Cloud Platform (GCP). • Support the client in the design of the architecture and the deploying the infrastructure of the continuous integration, continuous deployment and the continuous training of the Machine Learning models on Google Cloud Platform. • As Team Lead Data Scientist, I help in structuring, organizing, and monitoring data science projects and team. I also participate in the construction of pre-sales data-related solutions.

- Construction of a forecast model for invoiced turnover and EBITDA ; - Intelligent maintenance of large-scale sensors: Use of decision algorithms for predictive maintenance and optimization under constraint. This work is fully part of my work-study program at Cementys ; - Assistance in the development of an algorithm for counting and automatically calculating vehicle speed on the highways thanks to DAS technology.

Soutient scolaire en mathématiques

Mathematics tutor in bachelor's degree, First year level