一个 python + 数据预处理+随机森林模型 （案列）

pip install sklearn
pip install pandas
pip install joblib

import pandas as pd
import numpy as np
from sklearn import preprocessing

    ### 从Excel读
data_all=None
data_train=None
data_test=None
run_paint_or_ml=2


####################
# 1.2 数据预处理
####################
def run_preprocess(df,run_paint_or_ml):

    ###### 1) 舍弃不需要的列
    if run_paint_or_ml==1:
        ### 用于演示画图
        df=df.drop(['sysCustomerID'],axis=1)
        #df=df.drop(['sysCustomerID','sysContactID','ContactHrMi','Round','CampaignName','RemarkLength','Consider1','Consider9',],axis=1)
    else:
        ### 用于机器学习
        df=df.drop(['sysCustomerID','sysContactID','ContactDay','ContactHr','Team','Agent','ContactHrMi','Round','CampaignName','RemarkLength','Consider1','Consider9',],axis=1)

    ###### 2) 编码
    if run_paint_or_ml==1:
        ### 用于演示画图
        features = ['AgeGroup','Team','Agent',]
    else:
        ### 用于机器学习
        features = ['AgeGroup']

    for feature in features:
        le = preprocessing.LabelEncoder()
        le = le.fit(df[feature])
        df.is_copy =False
        df[feature] = le.transform(df[feature])

    print('\n==========> 预览预处理后的数据：==========')
    print(df.head())

    ###### 3) 分割数据集
    X=df.drop(['Result'],axis=1)
    y=df['Result']

    return df, X, y

# from sklearn.datasets import make_classification
from sklearn.metrics import make_scorer, accuracy_score, recall_score, precision_recall_fscore_support
from pprint import pprint

def scores(p,y):
    m = pd.DataFrame({'p':p,'y':y.values})
    tp = m[m.y+m.p>1].size/2
    pp = m[m.p>0].size/2
    ap = m[m.y>0].size/2
    print('true positive: {0}'.format(tp))
    print('predicted positive = true positive + false positive: {0}, tp/pp: {1}'.format(pp, tp/pp))
    print('actual positive = true positive + false negative: {0}, tp/ap: {1}'.format(ap, tp/ap))

    #print(m[m.y+m.p>0][0:m[m.y+m.p>0].size//2-1])
    '''
    i=0
    while i < m[m.y+m.p>0].size/2:
        print(m[m.y+m.p>0][i:min(i+50,int(m[m.y+m.p>0].size/2))])
        print('... ... ... ... ...')
        i+=50
    '''

####################
# 2.3 随机森林
####################
from sklearn.ensemble import RandomForestClassifier 
from sklearn.model_selection import GridSearchCV
import joblib
def run_RandomForest(X_train,y_train,X_test,y_test,threshold):
    print("-----------------随机森林----------------------------------")
    ###### 实例化模型
    clf=RandomForestClassifier()
    '''
    ###### 使用GridSearchCV调优参数
    ### 设置参数
    parameters={'n_estimators':[4,6,9],
                'criterion':['entropy','gini'],
                'max_features':['log2','sqrt','auto'],
                'max_depth':[2,3,5,10],
                'min_samples_leaf':[1,5,8],
                'min_samples_split':[2,3,5]
                }

    ### 测试参数
    acc_scorer=make_scorer(accuracy_score)
    gs=GridSearchCV(clf,parameters,scoring=acc_scorer,verbose=1)
    gs=gs.fit(X_train,y_train)

    ### 将最优参数赋给分类器
    print(gs.best_params_)
    clf=gs.best_estimator_
    '''
    ###### 直接使用调优得到的参数
    clf=RandomForestClassifier(criterion='entropy', max_depth=2, max_features='log2', min_samples_leaf=1, min_samples_split=2, n_estimators=4)

    ###### 加载模型
    #clf = joblib.load('rfc.pkl')

    ###### 训练数据
    clf.fit(X_train,y_train)

    ### 使用训练好的分类模型进行预测
    if threshold==1:
        predictions = clf.predict(X_test)
    else:
        # proba
        predictions = clf.predict_proba(X_test)[:,1]>threshold


    ###### 模型评估
    print('\n==========> 随机森林 --- precision：',accuracy_score(y_test, predictions))

  #这行代码使用了accuracy_score函数来计算随机森林模型的精确度。accuracy_score函数接受两个参数，
  #y_test表示测试集的真实标签，
  #predictions表示模型预测的标签。精确度是指模型预测为正例的样本中，真实为正例的比例。这行代码会输出随机森林模型的精确度。
  
    print('==========> 随机森林 --- recall：',recall_score(y_test, predictions))
  #这行代码使用了recall_score函数来计算随机森林模型的召回率。
  #recall_score函数也接受两个参数，
  #y_test表示测试集的真实标签，
  #predictions表示模型预测的标签。召回率是指真实为正例的样本中，模型预测为正例的比例。这行代码会输出随机森林模型的召回率**

  
    ### 打印出三个指标
    print(precision_recall_fscore_support(y_test, predictions,average='binary'))



    ###### 持久化
    ### 保存模型
    joblib.dump(clf, 'rfc.pkl')

    return predictions

def main():

    ### 从Excel读
    data_all=pd.read_csv('boci_trial_0423.csv')
    data_train=data_all.head(2000)
    data_test=data_all.tail(8000)
    #print('==========预览训练集：==========\n',data_all.head())

    ###### 1.2 数据预处理
    data_train, X_train, y_train = run_preprocess(data_train,run_paint_or_ml)
    data_test, X_test, y_test = run_preprocess(data_test,run_paint_or_ml)
    data_all, X_all, y_all = run_preprocess(data_all,run_paint_or_ml)
    ###### 2. 模型选择、训练、预测
    if run_paint_or_ml==2:

        #
        ###### 2.3 随机森林
        pred=run_RandomForest(X_train, y_train, X_test, y_test, 0.1)
        scores(pred,y_test)


if __name__=="__main__":
    main()

py sk.py