以不同类型的列作为训练数据集

Question

我之前只将一列（字符串类型数据）作为我的训练集，我想将另一个相应的列（浮点类型的金额列）与 Details 列一起考虑作为训练集。在金额栏中，负值表示借方，正值表示贷方。我该如何处理，我尝试将两列附加在一起，但我必须将 float 类型的金额转换为字符串类型，这不会在我的数据集中有任何意义。我想包括 Amount 列以检查机器是否可以学习变化，这在这种情况下非常重要。提前致谢。

Details                    |Amount               |Category
-------------------------------------------------------------                                
Tanishq Jwellery Bangalore |-990                 |jwellery
ODESK***BAL-28APR13        |240                  |Others
AEGON RELIGARE LIFE IN     |456                  |Others
INTERNET PAYMENT #999999   |-250                 |Transfer in for Card Payment
WWW.VISTAPRINT.IN          |245                  |Print
Khazana Jwellery           |-9000                |jwellery
INTERNET PAYMENT #999999   |785                  |Transfer in for Card Payment
Indian Oil                 |344                  |Fuel
Touch foot wear            |-782                 |Clothing

我的部分脚本：

import pandas as pd
import numpy as np
import scipy as sp
from sklearn.naive_bayes import MultinomialNB
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn import preprocessing
import time
import matplotlib.pyplot as plt  
from sklearn.model_selection import train_test_split 

# TRAIN DATA
data= pd.read_csv('ds1.csv', delimiter=',',usecols=['Details','Amount','Category'],encoding='utf-8')
data=data[data.Category !="Others"]

target_one=data['Category']
target_list=data['Category'].unique()

# TEST DATASET
test_data=pd.read_csv('ds2.csv', delimiter='\t',usecols=['Details','Amount','Category'],encoding='utf-8')

x_train, y_train = (data.Details, data.Category )
x_test, y_test = (test_data.Details, test_data.Category)

vect = CountVectorizer(ngram_range=(1,2))
X_train = vect.fit_transform(x_train)

X_test = vect.transform(x_test)
start = time.clock()

mnb = MultinomialNB(alpha =0.13)
mnb.fit(X_train,y_train)

result= mnb.predict(X_test)
print (time.clock()-start)

accuracy_score(result,y_test)

Answer 1

如果您只想将 "amount" 列堆叠到使用 CountVectorizer 获得的文本特征矩阵，只需在拟合 MultinomialNB:

之前执行此操作

import numpy as np

X_amount = data["Amount"].as_matrix().reshape(-1, 1)
X_train = X_train.toarray()
X_train = np.hstack((X_train, X_amount))
X_test_amount = test_data["Amount"].as_matrix().reshape(-1, 1)
X_test = X_test.toarray()
X_test = np.hstack((X_test, X_test_amount))

或者如果您想继续处理 X_train 的稀疏矩阵：

import scipy as sp

X_amount = data["Amount"].as_matrix().reshape(-1, 1)
X_train = sp.sparse.hstack((X_train, X_amount))
X_test_amount = test_data["Amount"].as_matrix().reshape(-1, 1)
X_test = sp.sparse.hstack((X_test, X_test_amount))

但是，我认为您最终会得到 ValueError: Input X must be non-negative，因为 MultinomialNB 旨在与 non-negative 特征值一起使用...

以不同类型的列作为训练数据集

Taking columns of different type as training dataset

python

machine-learning

scikit-learn

text-classification

data-science