tree+lr

main_tkz.py

import pandas as pd
from tree_tkz import tree_tkz
from logistic_regression_tkz import logistic_regression_tkz
from logistic_regression_test import logistic_regression_test
import numpy as np

if __name__ == '__main__':

    data = pd.read_table('../data/dz', header=None)
    data.fillna(0, inplace=True)

    data_columns = []

    data.columns = data_columns
    data['deal_type'] = np.where(data['deal_type'] > 0, 1, 0)

    data_1 = data[data['deal_type']==1]
    data = pd.concat([data_1, data.sample(frac=0.01, replace=True)])

    feature_columns = []

    data_train = data[feature_columns]
    tree_tkz(data_train, feature_columns)

    logistic_regression_tkz(data_train)

    data = pd.read_table('../data/dz', header=None)
    data.fillna(0, inplace=True)

    data.columns = data_columns
    data['deal_type'] = np.where(data['deal_type'] > 0, 1, 0)

    data_1 = data[data['deal_type'] == 1]
    data = pd.concat([data_1, data.sample(frac=0.01, replace=True)])

    data_test = data[feature_columns]

    logistic_regression_test(data_test)

tree_tkz.py

# -*- coding: utf-8 -*-

import pandas as pd
from sklearn import tree
import graphviz
import numpy as np
from re_tkz import re_tkz

#if __name__ == '__main__':
def tree_tkz(data, data_columns):

    #clf = tree.DecisionTreeRegressor()
    clf = tree.DecisionTreeClassifier()

    clf = clf.fit(data.drop(labels=data_columns[-1], axis=1), data[data_columns[-1]])

    feature_importances_list = clf.feature_importances_

    feature_importances_dict = {}

    for i in range(len(data_columns) - 1):
        feature_importances_dict[data_columns[i]] = feature_importances_list[i]

    print 'feature_importances_dict:', feature_importances_dict

    dot_data = tree.export_graphviz(clf, out_file=None,
                                    feature_names=data_columns[:-1],
                                    class_names=np.array(['1', '2']),
                                    filled=True, rounded=True,
                                    special_characters=True)
    graph = graphviz.Source(dot_data)
    graph.render("../data/tree")#view=True,显示pdf

    re_tkz()

re_tkz.py

# -*- coding: utf-8 -*-

import re
import json

#if __name__ == '__main__':
def re_tkz():

    with open('../data/tree', 'r') as f:
        tree_data = f.readlines()

    #<x1 &le; 1.5<br/>

    result = {}

    for data_line in tree_data:
        find_list = re.findall(r'<(.+?) &le; (.+?)<br/>', data_line)
        if find_list == []:
            continue
        t = find_list[0]
        if t[0] not in result.keys():
            result[t[0]] = []
        if float(t[1]) not in result[t[0]]:#去重
            result[t[0]].append(float(t[1]))
    for k in result.keys():#排序
        result[k].sort()
        result[k].append(9999999)
        result[k].insert(0, -9999999)
    with open('../data/cut_data.json', 'w') as f:
        json.dump(result, f)

logistic_regression_tkz.py

# -*- coding: utf-8 -*-

from sklearn.linear_model import LogisticRegression
from cut_data import cut_data
from sklearn.metrics import roc_auc_score
from sklearn.externals import joblib
from sklearn.model_selection import GridSearchCV


#if __name__ == '__main__':
def logistic_regression_tkz(data):

    x, y = cut_data(data)

    lr = LogisticRegression()

    '''
    parameters = {'penalty': ['l1', 'l2'], 'C': [1.0, 1.1, 1.2]}

    GSCV = GridSearchCV(lr, parameters, n_jobs=2)

    GSCV.fit(x, y)

    print GSCV.best_params_
    '''
    lr.fit(x, y)
    joblib.dump(lr, '../data/lr.model')


    print '权重:', lr.coef_#权重
    print '截距:', lr.intercept_#截距

    #y_pre = lr.predict_proba(x)
    y_pre = lr.predict(x)
    print 'train_auc:', roc_auc_score(list(y), list(y_pre))

cut_data.py

# -*- coding: utf-8 -*-

import pandas as pd
import json

def cut_data(data):

    with open('../data/cut_data.json', 'r') as f:
        cut_dict = json.load(f)

    for k in cut_dict.keys():
        #k_labels = [k + str(i) for i in range(len(cut_dict[k]) - 1)]
        data[k] = pd.cut(data[k], cut_dict[k])#labels = k_labels

    train_x = pd.get_dummies(data.drop(labels='deal_type', axis=1))
    train_y = data['deal_type']
    return train_x, train_y

logistic_regression_test.py

# -*- coding: utf-8 -*-

from cut_data import cut_data
from sklearn.metrics import roc_auc_score
from sklearn.externals import joblib

#if __name__ == '__main__':
def logistic_regression_test(data):

    x, y = cut_data(data)
    print x.shape
    lr = joblib.load('../data/lr.model')

    #y_pre = lr.predict_proba(x)
    y_pre = lr.predict(x)
    print 'test_auc:', roc_auc_score(list(y), list(y_pre))

转载于:https://www.cnblogs.com/kayy/p/9963491.html