You need to think about what you value the most out of: 1) accuracy 2) computation time & 3) interpret-ability for assessing model’s performance. 

We first have to decide what metric we can use to evaluate the performance of our model. Accuracy isn\’t always the best way to evaluate a model. It highly depends on whether we have balanced/imbalanced training dataset. For example, for classification models, precision, recall and f-score may be good measures instead of accuracy, especially in case of imbalanced data. Other measures are interpretability and speed.