gcForest 集成学习方法的 Python 实现

# -*- coding:utf-8 -*-
__author__ = 'Administrator'

from python_dailyworking.GCForest import gcForest
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from python_dailyworking.Get_KS import get_ks
import pandas as pd

#loading the iris data
iris = load_iris()
X = iris.data
Y = iris.target

X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.33)

gcf = gcForest(shape_1X=4, window = 2, tolerance=0.0)
gcf.fit(X_train, Y_train)

#predict 方法预测的是每一条样本的分类类别，结果 pred_X 是一个 [0,1,2...]的 array
pred_X = gcf.predict(X_test)
accuracy = accuracy_score(y_true=Y_test, y_pred=pred_X) #用 test 数据的真实类别和预测类别算准确率
print ('gcForest accuracy:{}'.format(accuracy))

#  predict_proba 方法预测的是每一条样本为 0，1,...类别的概率，结果是这样的：
# [[ 概率 1，概率 2，...],[ 概率 1，概率 2，...],...]的 DataFrame
# [:,1]表示取出序号为 1 的列，也就是预测类别为 1 的概率值,结果是一个数组
Y_predict_prod_test = gcf.predict_proba(X_test)[:, 1]
Y_predict_prod_train = gcf.predict_proba(X_train)[:, 1]
#下面是算 KS 值，不理解的可以置之不理
print ('model Y_test ks: ',get_ks(Y_test, Y_predict_prod_test))
print ('model Y_train ks: ',get_ks(Y_train, Y_predict_prod_train))

# -*- coding:utf-8 -*-
__author__ = 'Administrator'

import pandas as pd
import time
import numpy as np
from sklearn.model_selection import train_test_split
from python_dailyworking.GCForest import gcForest
from python_dailyworking.Get_KS import get_ks
from sklearn.metrics import accuracy_score

def train():
    input_path = 'D:\\working_temp_file\\final_combine_file_phone_xgb_77X.csv'
    input_data = pd.read_csv(input_path)
    drop_columns_all = 'bad'
    targetName = 'bad'
    X1 = input_data.drop(drop_columns_all, axis=1, inplace=False)
    X1 = X1.fillna(0)
    # 先将 X1 和 Y1（DataFrame）转化成 array
    X = np.array(X1)
    Y1 = input_data[targetName]
    Y1 = Y1.fillna(0)
    Y = np.array(Y1)

    X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.25)

    starttime = time.time()
    # gcForest Model
    # shape_1X 默认是 None，但是用到 Multi-Grain Scanning 时，需要用到，我理解的是特征的数量（对于序列数据）
    # window：滑动窗口的大小
    # stride：滑动步长，默认是 1
    # tolerance：提升的精度小于 tolerance 时，就停止整个过程。
    # n_mgsRFtree=30, cascade_test_size=0.2, n_cascadeRF=2,
    # n_cascadeRFtree=101, cascade_layer=np.inf,min_samples_mgs=0.1,
    # min_samples_cascade=0.05, tolerance=0.0, n_jobs=1
    gcf = gcForest(shape_1X=76, n_mgsRFtree=30, window=20,
                   stride=1, cascade_test_size=0.2, n_cascadeRF=2,
                   n_cascadeRFtree=50, cascade_layer=np.inf, min_samples_mgs=0.05,
                   min_samples_cascade=0.1, tolerance=0.0, n_jobs=1)
    gcf.fit(X_train, Y_train)

    pred_X = gcf.predict(X_test)
    print (pred_X)
    accuracy = accuracy_score(y_true=Y_test, y_pred=pred_X)
    print ('gcForest accuracy:{}'.format(accuracy))

    Y_predict_prod_test = gcf.predict_proba(X_test)[:, 1]
    Y_predict_prod_train = gcf.predict_proba(X_train)[:, 1]

    print ('model Y_test ks: ',get_ks(Y_test, Y_predict_prod_test))
    print ('model Y_train ks: ',get_ks(Y_train, Y_predict_prod_train))

    print ('整个模型的训练预测耗时为：',(time.time() - starttime))

if __name__ == '__main__':
    train()