train_word_vects = word_vects
train_classes = classes
train_dataset = [
get_doc_vector(words, vocabulary) for words in train_word_vects
]
# 训练贝叶斯模型
cond_probs, cls_probs = clf.train(train_dataset, train_classes)
# 测试模型
error = 0
for test_word_vect, test_cls in zip(test_word_vects, test_classes):
test_data = get_doc_vector(test_word_vect, vocabulary)
pred_cls = clf.classify(test_data, cond_probs, cls_probs)
if test_cls != pred_cls:
print('Predict: {} -- Actual: {}'.format(pred_cls, test_cls))
error += 1
print('Error Rate: {}'.format(error / len(test_classes)))
# 绘制不同类型的概率分布曲线
fig = plt.figure()
ax = fig.add_subplot(111)
for cls, probs in cond_probs.items():
ax.scatter(np.arange(0, len(probs)),
probs * cls_probs[cls],
label=cls,
alpha=0.3)
ax.legend()
test_word_vects.append(word_vects.pop(idx))
test_classes.append(classes.pop(idx))
train_word_vects = word_vects
train_classes = classes
train_dataset = [get_doc_vector(words, vocabulary) for words in train_word_vects]
# 训练贝叶斯模型
cond_probs, cls_probs = clf.train(train_dataset, train_classes)
# 测试模型
error = 0
for test_word_vect, test_cls in zip(test_word_vects, test_classes):
test_data = get_doc_vector(test_word_vect, vocabulary)
pred_cls = clf.classify(test_data, cond_probs, cls_probs)
if test_cls != pred_cls:
print('Predict: {} -- Actual: {}'.format(pred_cls, test_cls))
error += 1
print('Error Rate: {}'.format(error/len(test_classes)))
# 绘制不同类型的概率分布曲线
fig = plt.figure()
ax = fig.add_subplot(111)
for cls, probs in cond_probs.items():
ax.scatter(np.arange(0, len(probs)),
probs*cls_probs[cls],
label=cls,
alpha=0.3)
ax.legend()
train_word_vecs = word_vecs
train_labels = labels
train_dataset = [
get_doc_vector(words, vocabulary) for words in train_word_vecs
]
#训练贝叶斯模型
cond_probs, labels_probs = clf.train(train_dataset, train_labels)
#测试模型
error = 0
for test_word_vec, test_label in zip(test_word_vecs, test_labels):
test_data = get_doc_vector(test_word_vec, vocabulary)
pred_label = clf.classify(test_data, cond_probs, cls_probs)
if test_label != pred_label:
print("Predict:{} -- Actual:{}").format(pred_label, test_label)
error += 1
print("Error Rate:{}".format(error / len(test_labels)))
#绘制不同类型的概率分布曲线
fig = plt.figure()
ax = fig.add_subplot(111)
for label, probs in cond_probs.items():
ax.scatter(np.arange(0, len(probs)),
probs * label_probs[label],
label=label,
alpha=0.3)
ax.legend()
plt.show()
