机器学习的一些基本使用

绘图包和设置

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression

%matplotlib inline
plt.rcParams['font.sans-serif'] = 'SimHei'
plt.rcParams['axes.unicode_minus'] = False
%config Inlinebackend.figure_format = 'svg

连接sql

import numpy as np
import pandas as pd
from sqlalchemy import create_engine
from pyecharts.charts import *
from pyecharts import options as opts

engine = create_engine('mysql+pymysql://lzw:123456@localhost:3306/mydb?charset=utf8')
df = pd.read_sql('select * from db.house_info', engine)
df.to_sql('表名',engine,index=Flase,if_exists='replace')

#提取包含数据的字段
results = df['额外配件1'].str.contains('遥控')
results.fillna(value=False,inplace = True)
df[results]

#提取以什么开头的数据字段
df[df['industryField'].str.startswith('数据')]

'# 忽略警告
def ignore_warnings():
    warnings.filterwarnings('ignore')

#划分数据集

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test = train_test_split(data,target,test_size=0.2)

调参

from sklearn.model_selection import GridSearchCV

#svc模型
svc = SVC()

# 网格参数
param_grid={
    'C':[0.1,1,10],
    'gamma':[0.02,0.05,0.2]
}
# estimator=svc, 需要调参的模型
# param_grid=param_grid, 网格参数
# cv=5 默认是5折-None 训练组数
gv = GridSearchCV(estimator=svc,param_grid=param_grid,cv=5)

gv.fit(data,target)

# 最佳模型：带参数的模型
gv.best_estimator_

# 最佳参数
gv.best_params_

from sklearn.metrics import mean_squared_error
from sklearn.metrics import mean_absolute_error

# MSE 均方误差
mse_predict = mean_squared_error(y_test, y_predict)

# MAE 均方根误差
mae_predict = mean_absolute_error(y_test, y_predict)

# y_test：测试数据集中的真实值
# y_predict：根据测试集中的x所预测到的数值

字符串数字化（脱敏）

lis = data.columns
for c in lis:
    lis = df[f'{c}'].unique().tolist()
    dict1 = {i:v for i,v in zip(lis,range(1,len(lis)+1))}
    df[f'{c}'] = df[f'{c}'].map(dict1)

# 数据转换，将String/Object类型数据转换为int,用0,1,2,3...表示
for col in cols:
    col_unique=df[col].unique()
    df[col]=df[col].map(lambda x :np.argwhere(col_unique==x)[0,0])

最佳聚类参数

def best_n_clusters(data):
    scores=[]
    for k in range(2,20):
        #训练
        kmeans = KMeans(k)
        kmeans.fit(X)

        labels = kmeans.labels_
        score = silhouette_score(X,labels)
        scores.append(score)
    for i in range(len(scores)):
        if scores[i]==np.max(scores):
            return i+2