未在 AzureML 中获得预期的输出

Not Getting the Expected Output in AzureML

背景:我正在从事一个项目,旨在使用 Azure ML 中的情感分析将产品评论分为正面评论和负面评论。我在将评论分类到不同部门时卡住了。

我基本上是从 csv 文件中读取单词并检查评论(v:句子列表)是否包含这些单词。如果在评论中发现其中一些词,那么我会记下句子编号并将其推入相应的列表(FinanceList、QualityList、LogisticsList)。最后,我将列表转换为字符串并将它们推送到数据框中。

我在 Azure ML 脚本中编写的打印语句的输出未被记录。

数据框中的值总是变成 0,但是当我 运行 本地代码时,我得到了预期的输出。

第一张图片的描述:数据框的列显示 0 个值。

第二张图片的描述:我已经突出显示了我在本地获得的预期输出,用于 ​​AzureML 中使用的相同评论。

Image 1

Image 2

我已经检查过的东西:

  1. 已正确读取 csv 文件。
  2. 评论中包含我正在搜索的字词。

我无法理解哪里出错了。

'

import csv
import math
import pandas as pd
import numpy as np

def azureml_main( data, ud):

   FinanceDept = []
   LogisticsDept = []
   QualityDept = []
  #Reading from the csv files
   with open('.\Script Bundle\quality1.csv', 'rb') as fin:
      reader = csv.reader(fin)
      QualityDept = list(reader)

   with open('.\Script Bundle\finance1.csv', 'rb') as f:
      reader = csv.reader(f)
      FinanceDept = list(reader)

   with open('.\Script Bundle\logistics1.csv', 'rb') as f:
      reader = csv.reader(f)
      LogisticDept = list(reader)

   FinanceList = []
   LogisticsList = []
   QualityList = []

#Initializing the Lists   
   FinanceList.append(0)
   LogisticsList.append(0)
   QualityList.append(0)

   rev = data['Data']
   v = rev[0].split('.')

   print FinanceDept

   S = 0   
   for sentence in v:
      S = S + 1
      z = sentence.split(' ')
      for c in z:
         c = c.lower()
         if c in FinanceDept and S not in FinanceList:
            FinanceList.append(S)
         if c in LogisticsDept and S not in LogisticsList:
            LogisticsList.append(S)
         if c in QualityDept and S not in QualityList:
            QualityList.append(S)
   #Compute User Reputation Score
   Upvotes = int(ud['upvotes'].tolist()[0])
   Downvotes = int(ud['downvotes'].tolist()[0])
   TotalVotes = max(1,Upvotes+Downvotes)

   q = data['Score']

   print FinanceList

   repScore = float(Upvotes)/TotalVotes 
   repScore = repScore*float( q[0] )
   str1 = ','.join(str(e) for e in FinanceList) 
   str2 = ','.join(str(e) for e in QualityList)
   str3 = ','.join(str(e) for e in LogisticsList)

   x = ud['id']

   #df = pd.DataFrame(  [str(repScore), str1  , str2  , str3 ], columns=[Write the columns])
   d = {'id': x[0], 'Score': float(repScore),'Logistics':str3,'Finance':str1,'Quality':str2}
   df = pd.DataFrame(data=d, index=np.arange(1))
   return df,`

你能检查文件路径是否正确吗,根据 https://azure.microsoft.com/en-us/documentation/articles/machine-learning-execute-python-scripts/,Python 代码无法访问它运行的机器上的大多数目录,当前目录及其子目录除外-目录。 "Script Bundle" 是子目录吗?您也可以尝试使用输入而不是从脚本中读取 csv 文件。执行 Python 脚本模块接受 3 个输入,前两个是数据帧,第三个是为 python 库文件保留的。例如,您可以将输入 1 用于实际数据,将输入 2 用于最初在 csv 中的单词。需要有一种机制将 3 个 csv 文件捆绑到一个输入 2 的数据框中。

@Anuj Shankar, 经过同事测试,我们可以从CSV个文件中读取数据,得到了预期的结果。本经验请参考:

1) 输入数据 - 它有 apple.zip 文件,其中有两个 csv 文件与您相似,每个 csv 文件包含与公司相关的词袋。

2) Python 脚本:

# The script MUST contain a function named azureml_main
# which is the entry point for this module.
#
# The entry point function can contain up to two input arguments:
#   Param<dataframe1>: a pandas.DataFrame
#   Param<dataframe2>: a pandas.DataFrame
import csv
import numpy as np
import pandas as pd

def azureml_main(dataframe1 = None, dataframe2 = None):
    # Execution logic goes here
    #print('Input pandas.DataFrame #1:\r\n\r\n{0}'.format(dataframe1))

    # If a zip file is connected to the third input port is connected,
    # it is unzipped under ".\Script Bundle". This directory is added
    # to sys.path. Therefore, if your zip file contains a Python file
    # mymodule.py you can import it using:
    # import mymodule

    apple = {}
    microsoft = {}
  #Reading from the csv files
    with open('.\Script Bundle\apple.csv', 'rb') as f:
      reader = csv.reader(f)
      apple = list_to_dict(list(reader)[0])

    with open('.\Script Bundle\microsoft.csv', 'rb') as f:
      reader = csv.reader(f)
      microsoft = list_to_dict(list(reader)[0])

#    print('hello world' + ' '.join(apple[0]))
    applecount = 0
    microsoftcount = 0

    input = "i want to buy surface which runs on windows"
    splitted_input = input.split(' ')

    for word in splitted_input:
        if word in apple:
            applecount = applecount + 1
        if word in microsoft:
            microsoftcount = microsoftcount + 1

    print("apple bag of words count - " + str(applecount))
    print("microsoft bag of words count - " + str(microsoftcount))
    mydata = [{'input words': len(splitted_input)}, {'applecount':applecount},
        {'microsoftcount':microsoftcount}]       
    # Return value must be of a sequence of pandas.DataFrame
    return pd.DataFrame(mydata),


def list_to_dict(li):      
    dct = {}  
    for item in li:
        if dct.has_key(item):              
            dct[item] = dct[item] + 1  
        else:  
            dct[item] = 1  
    return dct  

3) 输出——如果我考虑字符串 "i want to buy surface which runs on windows"。它有 2 个与微软相关的词和 0 个与苹果相关的词,如下图所示。