How do you decide hyperparameters in the dataset


Please refer to given Link.

A good general way of tuning hyperparameters is a gridsearch. This will try numerous combinations of hyperparameters and rank them by an estimator.

So for example if you were doing a classification model you could use f1-score as your metric then add all the hyperparameters you think might affect the model into the parameter grid. Then loop through all the permutations in the grid and select the combination with the highest f1 as your optimized hyperparamers.

This works well on all datasets but has the benefit of being much quicker on a small dataset because the model fitting is typically quicker.