Management Solutions’ first Datathon, aimed at designing an Automated Machine Learning model capable of making good predictions on different datasets without modifying their code, was a success due to both the number of participants and the quality of the algorithms received
The Management Solutions Datathon was entered by 46 MS professionals with different backgrounds (physicists, mathematicians, engineers and economists) and from different geographies (Germany, Brazil, Chile, Colombia, Spain, United States and Peru) who, either individually or in pairs, designed 27 Auto ML algorithms.
Throughout the competition, participants had to face various choices regarding data processing, the choice of models and hyperparameter optimization, which was mostly done using grid searches - although some teams also used random search, genetic algorithms or Bayesian search.
The competition was very tight due to the high quality of the submitted algorithms, and ended with the victory of David Ara and Juan Martín Fornell who, as a prize, will attend the conference “London AI in Finance Summit” to be soon held in London.