【无标题】

官网数据集介绍：

数据集地址

link
在邀请某人约会或跳伞之前，了解你成功的可能性很重要。向潜在客户提供住房保险价格也是如此。Homesite是一家领先的房主保险提供商，目前没有一个动态转换率模型，可以让他们相信报价会导致购买
Homesite使用一个匿名的客户和销售活动信息数据库，包括财产和保险信息，让您预测哪些客户将购买给定的报价。准确预测转换将有助于Homesite更好地理解拟议定价变化的影响，并保持理想的客户细分组合。

文章使用模型

因为下述模型时间快并且正确率较高所以选择如下模型

model	accuary	time
KNN	0.7969637200406431	9.586912631988525
DTCclassifier	0.8685075608152532	15.190108060836792
RandomForest	0.8986910525372064	74.82114577293396

在正常串行编程中使用时间约为150s

串行代码

##导入数据

x_train = pd.read_table(r'./data/train_x.txt')
y_train = pd.read_table(r'./data/train_y.txt')
x_test = pd.read_table(r'./data/test_x.txt')
y_test = pd.read_table(r'./data/test_y.txt')

KNN

model_start = time.time()
knn_start = time.time()
KNNClassifier = KNeighborsClassifier(n_neighbors=6)
KNNClassifier.fit(x_train, np.ravel(y_train))
y_pred_KNN = KNNClassifier.predict(x_test)
KNNAcc = accuracy_score(y_pred_KNN, y_test)
knn_end = time.time()
print('KNN准确率为：{}'.format(KNNAcc))
print('KNN共耗时{}秒'.format(knn_end - knn_start))

决策树

DTCclassifier = DecisionTreeClassifier()
DTCclassifier.fit(x_train, np.ravel(y_train))
y_pred_DTC = DTCclassifier.predict(x_test)
DTCAcc = accuracy_score(y_pred_DTC, y_test)
print('决策树准确率为:{}'.format(DTCAcc))
print('决策树共耗时{}秒'.format(DTC_end - DTC_start))

随机森林

DTC_start = time.time()
DTCclassifier = DecisionTreeClassifier()
DTCclassifier.fit(x_train, np.ravel(y_train))
y_pred_DTC = DTCclassifier.predict(x_test)
DTCAcc = accuracy_score(y_pred_DTC, y_test)
DTC_end = time.time()
print('决策树准确率为:{}'.format(DTCAcc))
print('决策树共耗时{}秒'.format(DTC_end - DTC_start))

投票器

total = list(zip(y_pred_KNN, y_pred_DTC, y_pred_RF))
pre = [Counter(list(total[i])).most_common(1)[0][0] for i in range(len(total))]
vote = accuracy_score(pre, np.array(y_test).flatten())
print('投票:{}'.format(vote_acc))

模型并行

模型并行流程图

代码如下：

import pandas as pd
from sklearn.metrics import accuracy_score
from sklearn.neighbors import KNeighborsClassifier
import time
from collections import Counter
import numpy as np
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
import multiprocessing

# x_train = np.loadtxt('./data/train_x.txt', dtype=np.float32, delimiter='\t', skiprows=1)
# y_train = np.loadtxt('./data/train_y.txt', dtype=np.float32, delimiter='\t', skiprows=1)
# x_test = np.loadtxt('./data/test_x.txt', dtype=np.float32, delimiter='\t', skiprows=1)
# y_test = np.loadtxt('./data/test_y.txt', dtype=np.float32, delimiter='\t', skiprows=1)


x_train = pd.read_table('./data/train_x.txt')
y_train = pd.read_table('./data/train_y.txt')
x_test = pd.read_table('./data/test_x.txt')
y_test = pd.read_table('./data/test_y.txt')


def KNN_Model(x_train, y_train, x_test):
    KNNClassifier = KNeighborsClassifier(n_neighbors=5)
    KNNClassifier.fit(x_train, np.ravel(y_train))
    KNN = KNNClassifier.predict(x_test)
    KNNAcc = accuracy_score(KNN, y_test)
    data = open('KNN.txt', mode='w')
    for i in KNN.tolist():
        data.write(str(i) + '\n')
    print('最近邻准确率为：{}'.format(KNNAcc))
    return KNN


def DecisionTree_Model(x_train, y_train, x_test):
    DTCclassifier = DecisionTreeClassifier()
    DTCclassifier.fit(x_train, np.ravel(y_train))
    DTC = DTCclassifier.predict(x_test)
    DTCAcc = accuracy_score(DTC, y_test)
    data = open('DecisionTree.txt', mode='w')
    for i in DTC.tolist():
        data.write(str(i) + '\n')
    print('决策树准确率为：{}'.format(DTCAcc))
    return DTC


def RandomForest_Model(x_train, y_train, x_test):
    RFclassifier = RandomForestClassifier()
    RFclassifier.fit(x_train, np.ravel(y_train))
    randomforest = RFclassifier.predict(x_test)
    RFAcc = accuracy_score(randomforest, y_test)
    data = open('RandomForest.txt', mode='w')
    for i in randomforest.tolist():
        data.write(str(i) + '\n')
    print('随机森林准确率为:{}'.format(RFAcc))
    return randomforest


def load(path):
    pre = []
    for line in open(path, encoding='utf8'):
        line1 = line.replace("\n", "")
        pre.append(line1)
    return pre


def model(x_train, y_train, x_test):
    KNN_txt = load("./KNN.txt")
    DecisionTree_txt = load("./DecisionTree.txt")
    RandomForest_txt = load("./RandomForest.txt")
    total = list(zip(KNN_txt, DecisionTree_txt, RandomForest_txt))
    pre = [Counter(list(total[i])).most_common(1)[0][0] \
           for i in range(len(total))]
    vote = accuracy_score(pre, np.array(y_test).flatten())
    print('投票准确率为'.format(vote))


if __name__ == "__main__":
    model_start = time.time()
    task1 = multiprocessing.Process(target=KNN_Model, args=(x_train, y_train, x_test))
    task2 = multiprocessing.Process(target=DecisionTree_Model, args=(x_train, y_train, x_test))
    task3 = multiprocessing.Process(target=RandomForest_Model, args=(x_train, y_train, x_test))
    task1.start()
    task2.start()
    task3.start()
    task1.join()
    task2.join()
    task3.join()
    model_end = time.time()
    print('total_time{}'.format(model_end - model_start))

可提升方向

读写数据比较慢，模型加载比较慢这都是可以改进的地方