Hyperparameter Optimization inspired by MATLAB's global optimization pattern search algorythm. This python notebook contains some research I did on using Pattern Search to find the best hyperparameters for a machine learning algorythm. The tecnique looks promessing finding a better value with less function evaluations than scikit-opt Baysian Optimization and SKLearn Random search.
I also implemented sklearn's feature selector which automatically picks the most relevant features from a dataset.
The challenge is to predict sales and the data set is similar to the Rossman Sales Kaggle competition but for an Italian chain of retail stores.