import numpy as np
from sklearn.neighbors import NearestNeighbors, KNeighborsClassifier, RadiusNeighborsClassifier
import matplotlib.pyplot as plt
%matplotlib inline

samples = np.array(
                    [                        
                        ['SKU1', 0., 0.],
                        ['SKU2', 0.15, 0.3],
                        ['SKU3', 0.3, 0.6],
                        ['SKU4', 0.5, 0.7],                        
                        ['SKU5', 0.7, 0.65],
                        ['SKU6', 0.85, 0.7],
                        ['SKU7', 1., 1.],
                    ]
                   )

X = samples[:, 1:].astype(np.float32)
Y = samples[:, 0].astype(np.object)

radius = 0.2
point = np.array([[0.79, 0.80]])

#Вариант 1. Использование класса NearestNeighbors и метода kneighbors
nn = NearestNeighbors(n_neighbors=1, radius = radius, metric='euclidean', n_jobs=-1).fit(X)

distances, indices = nn.kneighbors(point)

print('Вариант 1')
print([Y[i] for i in np.asarray(indices)])


#Вариант 2 Использование класса NearestNeighbors и метода radius_neighbors
nn = NearestNeighbors(n_neighbors=1, radius = radius, metric='euclidean', n_jobs=-1).fit(X)

neigh_dist, neigh_ind = nn.radius_neighbors(point, sort_results = True)

print()
print('Вариант 2')
print([Y[i] for i in np.asarray(neigh_ind)])


#Вариант 3. Использование класса KNeighborsClassifier
knn = KNeighborsClassifier(n_neighbors=1, metric='euclidean', n_jobs=-1).fit(X, Y)
y_pred = knn.predict(point)

print()
print('Вариант 3')
print(y_pred)


#Вариант 4. Использование класса RadiusNeighborsClassifier
rnn = RadiusNeighborsClassifier(radius = radius, metric='euclidean', n_jobs=-1, ).fit(X, Y)
    
try:
    y_pred_ = rnn.predict(point)
except:
    y_pred_ = None

print()
print('Вариант 4')
print(y_pred_)

nn = NearestNeighbors(n_neighbors=1, radius = radius, metric='euclidean', n_jobs=-1).fit(X)