I build a model. Where I used KNN. I build a recommendation system with similarity measures feature of Knn. Here I calculate euclidean distance first, then calculate similarity(matching features/total features) and then calculate recommendation is (threshold < similarity)
Now my ques is, here I get accuracy almost 98-99% but my recall and f1_score is low. Like 10-12%
Is it normal? In recommendation tasks? I want to build a car recommendation

I tried a Knn model which measures similarity. My ques is does it measure f1_score, recall for total work?