Nijssen, SiegfriedCarvalho, AlissonAlissonCarvalho2025-02-042025-02-042021https://dial-mem.test.bib.ucl.ac.be/handle/123456789/26274During the development of a machine learning model, developers are required to make several decisions that will affect the results of this model. While they are tuning the hyperparameters of a model, they may choose one of several optimization techniques available to perform this task. The performance of this optimization technique, regarding how much it can improve the performance of the model, will directly depend on how suited this approach is to optimize this model. Today, most developers depend on their knowledge and expertise to choose which optimization method to employ in optimization tasks during the machine learning pipeline. This requires the developer to have good knowledge about the task being performed. To reduce the number of choices the developer will need to do during the machine learning pipeline while improving the results found, we propose a model that will predict which optimization method is the best suited for the current task, using previous experiences of the optimization techniques. This approach will profit from prior experiences to find the best optimization method to use in a task, diminishing the need of expertise a developer may need to solve a given task.AutoMLHyperparameter optimizationEvolutionary methodsMachine learningBayesian optimizationtext::thesis::master thesisthesis:33154