from sklearn import tree
def sklearn_tree(features, labels):
clf = tree.DecisionTreeClassifier()
clf.fit(features, labels)
return clf
if __name__ == '__main__':
from load_data import split_train_data,evaliate
tfs, tls, vfs, vls = split_train_data()
i = 0
mean_acc = 0
for (tfeatures,tlabels,vfeatures,vlabels) in zip(tfs, tls, vfs, vls):
print('tree : ', i, ' start')
clf = sklearn_tree(tfeatures, tlabels)
predict = clf.predict(vfeatures)
accuracy = evaliate(predict, vlabels)
mean_acc += accuracy
i += 1
print('tree ', i, ' :', accuracy)
print('mean acc : ', mean_acc / 10)
for (features, labels, vfeatures, vlabels) in zip(tfs, tls, vfs, vls):
print('net : ', i, ' train')
features = nd.array(features)
labels = nd.array(np.array(labels))
vfeatures = nd.array(np.array(vfeatures))
vlabels = nd.array(np.array(vlabels))
net = train(features, labels, './params/nn.params')
net = inter_output(net, 12)
features = net(features)
clf = sklearn_tree(features.asnumpy(), labels.asnumpy())
vfeatures = net(vfeatures)
predict = clf.predict(vfeatures.asnumpy())
print(net)
accuracy = evaliate(predict, vlabels.asnumpy())
print('accracy :', accuracy)
# predict = net(vfeatures)
# predict = predict_label(predict)
# i = i+1
# acc = accuracy(predict, vlabels)
# mean_acc += acc
# print('acc : ',acc)
break
print('mean acc : ', mean_acc / 10)
# train_features = nd.array(train_features,ctx=mx.cpu())
# train_labels = nd.array(train_labels,ctx=mx.cpu())
# net = train(train_features, train_labels)
# test_features = nd.array(test_features)
# print('start predict')
# predict = net(test_features)