Machine Learning that Helps Predict Customer Flow

Develop a model to predict the footfall, or rather the flow of customers, within different Vodafone stores. This was the task given by the Vodafone Challenge 2019 to the students of the Machine learning course, enrolled in the Bachelor in Economics, Management and Computer Science. "For each of the aound 300 Vodafone stores examined, we were been provided with 60 variables, such as store size, number of employees, location...," explains Giacomo Rebecchi, a student member of the winning group, made up of Alessandro Varnelli, Nikodem Jan Konarski, Waqas Zaheer, Rada Georgieva, and Emil Ferrari Herzum.

"We had to identify the variables to be included in the model chosen to calculate the footfall. We thought of including variables that had not been provided to us, such as the ratio between employees and the size of the stores, which we calculated through the others available data," continues Giacomo. "We decided to use more than one model and to provide the average of all of them for result of the forecast. It was a very important challenge for us: it is easier to understand this course when applying the theory to practical cases".

"We asked the students not only to use the machine learning methods learned in class in a creative way, but to go further, acquiring new methods and re-elaborating the data", says Riccardo Zecchina, professor and holder of the Vodafone Chair of Machine Learning and Data Science. "A strong stimulus to which the students reacted with great creativity and rigor: they surprised us by exceeding our expectations. A real lesson of what young people can do".