I have a Chaotic Bitcoin dataset with over 40+ variables and the response variable of Bitcoin price. I want to perform a feature selection on the dataset. I have tried various Filterbased, Embedded and Wrapper methods and finally zeroed down of feature selection using metaheuristic methods like PSO. Features selected using PSO are the closest that match with the existing literature. Could anybody, please explain if this is the right approach and what else I should try before making a final decision ?
#featureselection