knn_embedding.py

#lets assume an user ID 200
TEST_USER_ID = 200
#get the embedding of this user
user_embedding = user_model.predict([TEST_USER_ID]).reshape(1,-1)[0]

#create the KNN model
from sklearn.neighbors import KNeighborsClassifier
clf = KNeighborsClassifier(n_neighbors=11)
clf.fit(MOVIE_EMBEDDING_LIST, knn_train_label)

def recommend_movies(embedding):
    distances, indices = clf.kneighbors(embedding.reshape(1, -1),  n_neighbors=10)
    indices = indices.reshape(10,1)
    df_indices = pd.DataFrame(indices, columns = ['movie_id'])
    return df_indices.merge(movies,on='movie_id',how='inner',suffixes=['_u', '_m'])['title']
  
# get the recommendation
recommend_movies(user_embedding)