import numpy as np
import pandas as pd
import datetime

from sklearn.impute import SimpleImputer #预处理模块
from sklearn.model_selection import  train_test_split  #训练集和测试集模块
from sklearn.metrics import classification_report #预测结果评估模块

from sklearn.neighbors import  KNeighborsClassifier #K近邻分类器
from sklearn.tree import DecisionTreeClassifier #决策树分类器
from sklearn.naive_bayes import GaussianNB #高斯朴素贝叶斯函数

starttime = datetime.datetime.now()

def load_datasets(feature_paths, label_paths):
    feature = np.ndarray(shape=(0, 41)) #列数量和特征维度为41
    label = np.ndarray(shape=(0, 1))
    for file in feature_paths:
        #逗号分隔符读取特征数据，将问号替换标记为缺失值，文件不包含表头
        df = pd.read_table(file, delimiter=',', na_values='?', header=None)
        #df = df.fillna(df.mean()) #若SimpleImputer无法处理nan，则用pandas本身处理
        #使用平均值补全缺失值，然后将数据进行补全
        imp = SimpleImputer(missing_values=np.nan, strategy='mean') #此处与教程不同，版本更新，需要使用最新的函数填充NAn，暂不明如何调用
        imp.fit(df) #训练预处理器 此句有问题
        df = imp.transform(df) #生产预处理结果
        feature = np.concatenate((feature, df))#将新读入的数据合并到特征集中

    for file in label_paths:
        df = pd.read_table(file, header=None)
        #将新读入的数据合并到标签集合中
        label = np.concatenate((label, df))
    #将标签归整为一维向量
    label = np.ravel(label)
    return feature, label

if __name__ == '__main__':
    #读取文件，根据本地目录文件夹而设定
    path = 'D:\python_source\Machine_study\mooc_data\classification\dataset/'
    featurePaths, labelPaths = [], []
    for i in range(0, 5, 1):  #chr(ord('A') + i)==B/C/D
        featurePath = path + chr(ord('A') + i) + '/' + chr(ord('A') + i) + '.feature'
        featurePaths.append(featurePath)
        labelPath = path + chr(ord('A') + i) + '/' + chr(ord('A') + i) + '.label'
        labelPaths.append(labelPath)
    #将前4个数据作为训练集读入
    x_train, y_train = load_datasets(featurePaths[:4], labelPaths[:4])
    #将最后一个数据作为测试集读入
    x_test, y_test = load_datasets(featurePaths[4:], labelPaths[4:])
    #使用全量数据作为训练集，借助函数将训练数据打乱，便于后续分类器的初始化和训练
    x_train, x_, y_train, y_ = train_test_split(x_train, y_train, test_size=0.0)

    print('Start training knn')
    knn = KNeighborsClassifier().fit(x_train, y_train)    #使用KNN算法进行训练
    print('Training done')
    answer_knn = knn.predict(x_test)

    print('Start training DT')
    dt = DecisionTreeClassifier().fit(x_train, y_train)   #使用决策树算法进行训练
    print('Training done')
    answer_dt = dt.predict(x_test)
    print('Prediction done')

    print('Start training Bayes')
    gnb = GaussianNB().fit(x_train, y_train)    #使用贝叶斯算法进行训练
    print('Training done')
    answer_gnb = gnb.predict(x_test)
    print('Prediction done')

    #对分类结果从 精确率precision 召回率recall f1值fl-score和支持度support四个维度进行衡量
    print('\n\nThe classification report for knn:')
    print(classification_report(y_test, answer_knn))
    print('\n\nThe classification report for DT:')
    print(classification_report(y_test, answer_dt))
    print('\n\nThe classification report for Bayes:')
    print(classification_report(y_test, answer_gnb))
    endtime = datetime.datetime.now()
    print(endtime - starttime) #时间统计