An important airport management company entrusted us with a project related to the analysis and understanding of passenger behavior.
The challenge is to use machine learning techniques to predict passenger traffic and support the planning of airport staff at Points of Interest in Airport Areas according to expected passenger traffic.
An ML solution has been created following these 4 objectives, fundamental to correctly address the issue of Forecast:
– Follow the change of data over time and update quickly
– Have the availability of temporal data
– Follow a process that follows the principle of AutoML
– Provide an end-to-end ML service oriented to Cloud services
From a technical point of view, we have defined new optimization criteria for Time Series, through a protocol developed in Humanativa that applies training criteria on mobile time windows but aims to:
– Simplify and automate model lifecycle management.
– Ensure performance monitoring.
– Improve the quality of production models.
– Improve governance with MLOps services.
From a product point of view, this initiative like others of Forecast has led us to the development of an AutoForecast engine making processes of this type increasingly automatable and reusable of Customer Analysts Experts of the Domain.
Realization of the end-to-end Machine Learning service, from the creation of Machine Learning Models, to MLOPS Procedures and Advanced Analytics production.