Question about cadets dataset results

[m-shayan73]

Hello, I am seeing a difference in results I get from the provided "distance_save_cadets.pkl" in the repo and the one I create from the provided code.

Results I get when I run eval.py with the provided "distance_save_cadets.pkl" file:

Results I get if i create a new distance_save_cadets.pkl" file from the provided code:

There is a notable increase in the false positives (and decrease in true negatives hence a dip in precision and f1 score). For both, I am using the same/default condition to find best_idx (if dataset == 'cadets' and rec[i] < 0.9976) and n_neighbors (200).

Am I missing something? Would appreciate any help.