ccc-机器学习算法基础-3

KNN API（预测入住位置）

from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import  train_test_split
from sklearn.preprocessing import StandardScaler
import pandas as pd


def knncls():
    data = pd.read_csv("D:/BaiduNetdiskDownload/train.csv")
	#缩小数据范围
    data = data.query("x > 1.0 & x <1.25 & y > 2.5 & y < 2.75")
    #处理时间戳格式
    time_value = pd.to_datetime(data['time'],unit='s')
    time_value = pd.DatetimeIndex(time_value)
    data["day"] = time_value.day
    data["hour"] = time_value.hour
    data["weekday"] = time_value.weekday
    data = data.drop(['time'],axis=1)
    
    place_count = data.groupby('place_id').count()
	#继续缩小范围
    tf = place_count[place_count.row_id > 30].reset_index()
    data = data[data['place_id'].isin(tf.place_id)]
    data = data.drop(['row_id'], axis=1)
    y = data['place_id']
    x = data.drop(['place_id'],axis=1)
    x_train, x_test, y_train, y_test = train_test_split(x,y,test_size=0.25)
	#标准化，提高准确率
    std = StandardScaler()
    x_train = std.fit_transform(x_train)
    x_test = std.transform(x_test)
	#调用API
    knn = KNeighborsClassifier(n_neighbors=5)
    knn.fit(x_train,y_train)
    y_predict = knn.predict(x_test)
    print("*****************")
    print("预测的目标签到位置：",y_predict)
    print("*****************")
    print("预测的准确率:",knn.score(x_test,y_test))
    return None

if __name__ == "__main__":
    knncls()

说明：

• 时间戳类型的数据可以使用pandas中的to_datetime处理后再用DatetimeIndex分割
• 提高准确率的方式：
  1.删除无关的列，如用于筛选的row_id
  2.标准化
  3.调整K值
  4.排除异常值
• K值过小容易受异常点影响，K值过大容易受K值的数量波动影响
• 对计算机的计算消耗较大

Bayes API (文章分类)

from sklearn.datasets import fetch_20newsgroups
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.model_selection import  train_test_split
from sklearn.naive_bayes import MultinomialNB

def naviebayes():
    news = fetch_20newsgroups(subset='all')
    x_train, x_test, y_train, y_test = train_test_split(news.data,news.target)

    tf = TfidfVectorizer()
    x_train = tf.fit_transform(x_train)
    x_test = tf.transform(x_test)

    mlt = MultinomialNB(alpha=1.0)
    mlt.fit(x_train, y_train)
    y_predict = mlt.predict(x_test)
    print("预测文章类别为:",y_predict)
    print("准确率:",mlt.score(x_test,y_test))



if __name__ == '__main__':
    naviebayes()

说明：

• 朴素贝叶斯准确率主要与训练集有关，不容易通过调参得到比较好的提升
• MultinomialNB()参数说明
• 平滑系数用来防止0结果的产生
• 贝叶斯公式
  
  P(F1,F2,…) 预测文档中每个词的概率
  P（C)：每个文档类别的概率(某文档类别词数／总文档词数)

精确率和召回率

混淆矩阵 （多分类）

精确率：预测结果为正例样本中真实为正例的比例

召回率：真实为正例的样本中预测结果为正例的比例

F1-score：模型的稳健型

n折交叉验证：n等分所有数据

多次测试，取平均值

改良KNN（超参数搜索-网格搜索 Grid API）

当参数很过时，手动调整过程繁杂，所以需要对模型预设几种超参数组合。每组超参数都采用交叉验证来进行评估。最后选出最优参数组合建立模型。

from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import  train_test_split,GridSearchCV
from sklearn.preprocessing import StandardScaler

import pandas as pd


def knncls():
    data = pd.read_csv("D:/BaiduNetdiskDownload/train.csv")
    data = data.query("x > 1.0 & x <1.25 & y > 2.5 & y < 2.75")
    time_value = pd.to_datetime(data['time'],unit='s')
    time_value = pd.DatetimeIndex(time_value)
    data["day"] = time_value.day
    data["hour"] = time_value.hour
    data["weekday"] = time_value.weekday
    data = data.drop(['time'],axis=1)

    place_count = data.groupby('place_id').count()
    tf = place_count[place_count.row_id > 30].reset_index()
    data = data[data['place_id'].isin(tf.place_id)]
    data = data.drop(['row_id'], axis=1)
    y = data['place_id']
    x = data.drop(['place_id'],axis=1)
    x_train, x_test, y_train, y_test = train_test_split(x,y,test_size=0.25)

    std = StandardScaler()
    x_train = std.fit_transform(x_train)
    x_test = std.transform(x_test)

    knn = KNeighborsClassifier()
    param = {"n_neighbors":[3,5,10]}
    gc = GridSearchCV(knn,param_grid=param,cv=2)
    gc.fit(x_train,y_train)

    print("在测试集上的准确率:",gc.score(x_test,y_test))
    print("在交叉验证中最好的结果:",gc.best_score_)
    print("最好模型:",gc.best_estimator_)
    print("每个超参数每次交叉验证的结果:",gc.cv_results_)
    return None

if __name__ == "__main__":
    knncls()

说明：

• GridSearchCV()参数说明：