Humberto started his research creating new algorithms for Combinatorial Optimization Problems, including heuristics and exact solutions. His results are reference in the Vehicle Routing Problem with Time Windows and in its dynamic version.
In 2009, he began a new phase in his career. The projects were focused in new algorithms for financial markets.
At the beginning, the attempts were related to predict intraday returns. After several attempts, he changed the focus to algorithm trading. His conclusion was that you can create some predictors better than random algorithms but you need of a good system to communicate with financial markets.
In this context, Humberto modeled a Simulator for Brazilian Financial Market to be used as base in the research experiments. Then, in 2011 he and his research group made a first version.
In 2012, during a technical visit in a Brazilian company, they realized that the old simulator had several differences with the real life problem. So, they made a new simulator based on the Fix Protocol. In 2013, the simulator was done and the research group started the experiments with attempts of different Algotraders. This simulator can be seen here.
Since 2013, he has worked trying to create new Market Makers Strategies using Artificial intelligence and Optimization techniques.
During this evolution, Humberto realized that the majority of market maker’s problems is related with Binary Classification, but with a very complex scenario: the data sets are imbalanced. So, since 2015, Humberto and his research group have studied this specific problem using different approaches.
During his career, dozens of people studied at the Research & Development Lab. Some of them are professors, Ph.D. candidates, M.Sc. candidates in many universities.