Hyper tuning reduce the accuracy score, why?

I have performed hyper tuning grid CV search on KNN model. The actual accuracy score for my KNN was accuracy of 42.31 % without performing hyper tuning. However, after performing hyper tuning, the accuracy score has gone down to 27.97 %? I would like to know why this happened? I performed Near Miss (under sampling) to balance my dataset.

• So you had an initial $K$ with accuracy of $42.31\%$ accuracy and have evaluated different $K$, where the best $K^{*}$ resulted in $27.97\%$ accuracy ? Sep 19 '20 at 10:53
• We need more details. What set are these scores reported for? What's the hyperparameter grid, the default hyperparameters, the selected hyperparameters? When in your pipeline do you undersample? Sep 19 '20 at 14:05
• Try to make sure that the default values of parms are part of your parm-values dict if that parm is in the dict. e.g. n_neighbors=5, *, weights='uniform', algorithm='auto', leaf_size=30, p=2, metric='minkowski', metric_params=None Sep 20 '20 at 8:30