朴素贝叶斯分类器（姓名预测性别）

使用的数据集包含两列，name(姓名)，sex(性别), 数量45000, name列数据唯一。 代码实现：

import random
import nltk
import pandas as pd
from pathlib import Path
from sklearn import model_selection
from numpy import mean

current_path = Path.cwd()

# 特征提取
def gender_features(name):
    name = name.lower()
    if len(name) == 2:
        return {
            'last_name': name[-1]
        }
    if len(name) >= 3:
        return {
            'last_name': name[-1],
            'last2_name': name[-2],
            'last12_name': name[-2:]
        }

# 获取featuresets
def get_featuresets(X, y):
    labeled_names = []
    for i in range(len(X)):
        labeled_names.append((X.values[i], y.values[i]))

    # 数据打乱
    random.shuffle(labeled_names)
    # 我们使用特征提取器来处理数据
    featuresets = [(gender_features(name), gender) for (name, gender) in labeled_names]
    return  featuresets

if __name__=='__main__':
    labeled_names = []
    df = pd.read_csv(Path(current_path, '中文姓名性别预测.csv'), encoding='utf8')
    # K折交叉验证
    kf = model_selection.KFold(n_splits=10)
    # 使用10折交叉验验证划分数据集，返回一个生成器对象（即索引）
    digits_gen = kf.split(df)

    accuracy_list = []
    for train_idx, test_idx in digits_gen:
        X_train = df['name'].iloc[train_idx] #训练集
        X_test = df['name'].iloc[test_idx] #测试集
        y_train = df['sex'].iloc[train_idx] #训练集标签
        y_test = df['sex'].iloc[test_idx] #测试集标签

        featuresets_train = get_featuresets(X_train, y_train)
        featuresets_test = get_featuresets(X_test, y_test)

        # 该训练集用于训练一个新的“naive Bayes”分类器。
        classifier = nltk.NaiveBayesClassifier.train(featuresets_train)
        accuracy_list.append(nltk.classify.accuracy(classifier, featuresets_test))
    print(accuracy_list)
    print(mean(accuracy_list))