I have used
nbrs = NearestNeighbors(metric= 'cosine', algorithm='brute').fit(items_features)
distances, indices = nbrs.kneighbors(item_features)
to find some suggested items based on some attributes in the dataset. I really find the suggested items but I need to see if these items are really the more nearest items.
I tried this function to find the similarity between distances of A with each items B or C or D spatial.distance.cosine(distances of A,distances of B/C/D)
One output for my code is
B with 0.0004864387565900463
C with 0.003675794744087746
D with 0.05855020953442236
E with 0.0048087457894999686
I thought the items must be orderd from the neighrest but E is break the rule. Is those distances in above is logic or there is a wrong.
How can I make sure my algorithm work well???
