def score(self, X, y):
"""
check the performance of the model;
@X: NxD numpyarray, the data whose each row is an example;
@y: 1D numpyarray, labels;
"""
y_pred, _ = self.predict(X)
y_pred[y_pred == -1] = 0
accuracy = pm.accuracy(y, y_pred)
precision = pm.precision(y, y_pred)
recall = pm.recall(y, y_pred)
Fmeasure = pm.F_measure(y, y_pred)
return (accuracy, precision, recall, Fmeasure)
def oobTest(X, y, oobs, forest):
"""
use the out of bags to test the randorm forest classifier;
@X: NxD numpyarray, the data whose each row is an example;
@y: 1D numpyarray, labels;
@oobs: list, a 2D list to store the list of oob;
@forest: list, store the trees
"""
# for binary classification
if len(np.unique(y)) == 2:
accuracy = 0
precision = 0
recall = 0
Fmeasure = 0
for i in range(len(forest)):
X_i = X[oobs[i]]
y_i = y[oobs[i]]
y_pred = predict(X_i, forest)
accuracy += pm.accuracy(y_i, y_pred)
recall += pm.recall(y_i, y_pred)
precision += pm.precision(y_i, y_pred)
Fmeasure += pm.F_measure(y_i, y_pred)
return (accuracy/len(forest), precision/len(forest),
recall/len(forest), Fmeasure/len(forest))
else:
accuracy = 0
for i in range(len(forest)):
X_i = X[oobs[i]]
y_i = y[oobs[i]]
y_pred = predict(X_i, forest)
accuracy += pm.accuracy(y_i, y_pred)
return accuracy/len(forest)
accuracy = 0
precision = 0
recall = 0
F_measure = 0
for train, val in kfold.split(data):
X_train = data[train, :-1]
y_train = data[train, -1]
X_val = data[val, :-1]
y_val = data[val, -1]
knn = KNearestNeighbors()
knn.train(X_train, y_train)
y_pred = knn.predict(X_val, k+1, "l2")
accuracy += pm.accuracy(y_val, y_pred)
precision += pm.precision(y_val, y_pred)
recall += pm.recall(y_val, y_pred)
F_measure += pm.F_measure(y_val, y_pred)
accuracy = accuracy/10
precision = precision/10
recall = recall/10
F_measure = F_measure/10
accuracy_hist[k] = accuracy
precision_hist[k] = precision
recall_hist[k] = recall
F_measure_hist[k] = F_measure
'''
import PCA
data_2dim = PCA.PCA(data, 2)
data_df = pd.DataFrame(data_2dim, index = data[:,-1])