Should I choose higher AUC over higher accuracy

I have imbalanced classification problem. I did some method such as oversampling using random oversampling and SMOTE-TOMEK. In random oversampling, it has higher AUC but lower accuracy compared to SMOTE-TOMEK. The accuracy after random oversampling is below 0.7 while the accuracy after using SMOTE-TOMEK is above 0.7. Both AUC are above 0.5. Should I prefer random oversampling over SMOTE-TOMEK although the accuracy in random oversampling is lower than SMOTE-TOMEK?