def test_weights(self):
"""Testea los weights"""
# Primero, caso en el que considere todos los vecinos, y en caso
# de empate deberia dar el primero.
classifier = metnum.KNNClassifier(12, "uniform")
X_train = np.array([
[20, 20],
[21, 20],
[20, 22],
[22, 20],
[0, 0],
[1, 0],
[0, 1],
[1, 1],
[5, 5],
[6, 5],
[5, 6],
[6, 6],
])
y_train = np.array([
[9],
[9],
[9],
[9],
[0],
[0],
[0],
[0],
[5],
[5],
[5],
[5],
])
classifier.fit(X_train, y_train)
X = np.array([
[15, 15], # deberia dar 9
])
got = classifier.predict(X)
assertAllClose(self, np.array([0.0]), got)
# prediciendo tomando como weights la distancia,
# deberia dar el valor correcto
classifier = metnum.KNNClassifier(12, "distance")
classifier.fit(X_train, y_train)
got = classifier.predict(X)
assertAllClose(self, np.array([9.0]), got)
def run_no_pca(dfTrain, dataTest):
dataTrain = dfTrain[dfTrain.columns[1:]].values
labelTrain = dfTrain["label"].values.reshape(-1, 1)
dataTest = dataTest.values
clf = metnum.KNNClassifier(k)
clf.fit(dataTrain, labelTrain)
pred = clf.predict(dataTest)
return pred
def run_pca(dfTrain, dataTest):
dataTrain = dfTrain[dfTrain.columns[1:]].values
labelTrain = dfTrain["label"].values.reshape(-1, 1)
dataTest = dataTest.values
pca = metnum.PCA(alpha)
pca.fit(dataTrain)
dataTrain = pca.transform(dataTrain)
dataTest = pca.transform(dataTest)
clf = metnum.KNNClassifier(k)
clf.fit(dataTrain, labelTrain)
pred = clf.predict(dataTest)
return pred
def test_little(self):
"""Testea un caso en R2"""
classifier = metnum.KNNClassifier(3, "uniform")
X_train = np.array([
[0, 0],
[1, 0],
[0, 1],
[1, 1],
[5, 5],
[6, 5],
[5, 6],
[6, 6],
[20, 20],
[21, 20],
[20, 22],
[22, 20],
])
y_train = np.array([
[0],
[0],
[0],
[0],
[5],
[5],
[5],
[5],
[9],
[9],
[9],
[9],
])
classifier.fit(X_train, y_train)
X = np.array([
[15, 15], # 9
[10, 10], # 5
[2, 2], # 0
])
want = np.array([9.0, 5.0, 0.0])
got = classifier.predict(X)
assertAllClose(self, want, got)
def test_same(self):
"""Testea que predecir la matriz de entrenamiento de el mismo res"""
classifier = metnum.KNNClassifier(1, "uniform")
X_train = np.array([
[1, 2, 3],
[6, 6, 6],
[0, 1, 0],
])
y_train = np.array([
[0],
[2],
[3],
])
classifier.fit(X_train, y_train)
y = classifier.predict(X_train)
self.assertTrue(np.allclose(y.reshape(3, 1), y_train))