对离散特征建子模型实验

 

 

 

In [1]:

# coding=utf8
%matplotlib inline
import numpy as np
import pandas as pd
from sklearn.ensemble import RandomForestClassifier as rf
from sklearn.linear_model import LogisticRegression as lr
import matplotlib.pyplot as plt
from sklearn.cross_validation import train_test_split
from sklearn.metrics import accuracy_score

def show_importance_of_sub_model(total_size,factor_size,special_percent):
    each_dt_size = total_size/factor_size
    dt = pd.DataFrame()
    y = pd.DataFrame()
    special_size = int(factor_size*special_percent)
    special = range(special_size)

    for i in range(factor_size):
        dt_current = pd.DataFrame(np.random.random((each_dt_size,5)))
        dt_current['type'] = 'type'+str(100+i)
        y_current = pd.DataFrame(np.ones((each_dt_size,)))

        if i in special:
            y_current[(dt_current[0]+dt_current[1]+dt_current[2])>0.6] = -1
        else:
            y_current[(dt_current[0]+dt_current[1]+dt_current[2])>1.5] = -1

        y = pd.concat((y,y_current))
        dt = pd.concat((dt,dt_current))

    type = pd.get_dummies(dt['type'])
    y = np.array(y[0])

    lr_clf = lr(penalty='l1')

    n_estimator = 100

    dummy_performance = []
    sub_model_performance = []
    for i in range(10):
        dt_train,dt_test,y_train,y_test,type_train,type_test = train_test_split(dt,y,type,test_size=0.2)
        X_ori_train =  pd.concat((dt_train,type_train),1)
        del X_ori_train['type']
        X_ori_test =  pd.concat((dt_test,type_test),1)
        del X_ori_test['type']

        lr_clf.fit(type_train,y_train)
        coef = lr_clf.coef_[0]

        dt_train.loc[:,'coef'] = 0
        dt_test.loc[:,'coef'] = 0
        for i in range(len(type.columns)):
            dt_train.loc[dt_train['type']==type.columns[i],'coef'] = coef[i]
            dt_test.loc[dt_test['type']==type.columns[i],'coef'] = coef[i]
        X_sub_train = dt_train
        del X_sub_train['type']
        X_sub_test = dt_test
        del X_sub_test['type']

        clf_ori = rf(n_estimators=n_estimator)
        clf_ori.fit(X_ori_train,y_train)
        dummy_performance.append(accuracy_score(y_test,clf_ori.predict(X_ori_test)))

        clf_sub = rf(n_estimators=n_estimator)
        clf_sub.fit(X_sub_train,y_train)
        sub_model_performance.append(accuracy_score(y_test,clf_sub.predict(X_sub_test)))
    return [np.mean(dummy_performance),np.mean(sub_model_performance)]

def performance_ev(factor_size=100,special_percent=0.2,total_size=10000):
    x = None
    outcome = list()
    if type(factor_size) == list:
        x = factor_size
        factor_sizes = factor_size
        for factor_size in factor_sizes:
            outcome.append(show_importance_of_sub_model(total_size,factor_size,special_percent))
    else:
        x = special_percent
        special_percents = special_percent
        for special_percent in special_percents:
            outcome.append(show_importance_of_sub_model(total_size,factor_size,special_percent))

    performance = pd.DataFrame(outcome,columns=['dummy','sub_model'])
    plt.plot(x,performance)

In [2]:

%matplotlib inline
performance_ev(factor_size = [5,10,20,40,80])

 

In [3]:

%matplotlib inline
performance_ev(special_percent = [0.02,0.04,0.08,0.16,0.32])

 

 

对离散特征，使用子模型建模能提升效果

离散特征越重要(重要的值越多，对应special_percent越高)，使用子模型带来的提升越大

离散特征的取值越多，RandomForest的预测效果越差