A safer, more efficient and interactive version of the Portal Bioinfo (https://bioinfo.lncc.br/) is being implemented by applying decision tree techniques for the construction of predictive models using Orange machine learning programs and scikit-learn that use data from provenance and execution of bioinformatics applications in the Santos Dumont supercomputer. The creation of prediction models will provide a more efficient use of the computational and parallel features of the supercomputer, as the predictions will be used to optimize the configuration of the use of clusters (type and quantity) for executions to achieve greater efficiency (using as cut-off 75%) which will allow better disposition and use of the SDumont supercomputer environment when the machines are allocated to a given experiment.