Skip to content
Related Articles

Related Articles

Improve Article

How to convert unstructured data to structured data using Python ?

  • Last Updated : 12 Jul, 2021

Prerequisite: What is Unstructured Data?

Sometimes machine generates data in an unstructured way which is less interpretable. For example, Biometric Data, where an employee does Punch – IN or OUT several times with mistakes. We can not analyze the data and identify the mistakes unless it’s in a tabular form. In this article, we will take unstructured biometric data and convert it into useful information in terms of a table.


Here we will work with Daily Punch – In Report. Data is given below. Punch records captured for Main Door and Second Door. Main Door is Outdoor Gate and Second Door is Project Room Gate. We need to identify which employee spent how much time in the project room or second door. The dataset we’re going to use is Bio.xlsx:

 Here is the Biometric Data for John Sherrif where punch in and punch out records has given for Main Door and Second Door.

Emp Code:  COMP123:John Sherrif
Att. DateStatusPunch Records   

08:33:in(Second Door),08:35:(Second Door),08:37:(Main Door),09:04:out(Second Door),09:09:in(Second Door),

09:15:out(Second Door),09:15:(Second Door),09:18:(Second Door),09:52:in(Second Door),09:54:(Second Door),

10:00:out(Main Door),10:17:in(Main Door),10:53:out(Second Door),11:47:in(Second Door),11:47:(Second Door),

11:49:(Second Door),11:50:(Second Door),13:08:out(Second Door),13:09:(Second Door),13:12:(Second Door),

13:14:in(Second Door),13:36:out(Second Door),13:36:(Second Door),14:27:in(Second Door),14:32:out(Main Door),

14:48:in(Second Door),14:48:(Second Door),14:49:(Second Door),14:52:(Main Door),14:56:out(Second Door),

14:57:(Second Door),14:59:(Second Door),15:04:in(Second Door),16:22:out(Second Door),16:34:in(Second Door),

19:58:out(Main Door),  

The above Data is not insightful to analyze. Our desired output is:

Emp Code                                             Punch – IN                                           Punch – OUT                                                           
COMP123:John Sherrif08:33:in(Second Door)09:04:out(Second Door)
COMP123:John Sherrif09:09:in(Second Door)09:15:out(Second Door)
COMP123:John Sherrif09:52:in(Second Door)10:53:out(Second Door)
COMP123:John Sherrif11:47:in(Second Door)13:08:out(Second Door)
COMP123:John Sherrif13:14:in(Second Door)13:36:out(Second Door)
COMP123:John Sherrif14:27:in(Second Door)out
COMP123:John Sherrif14:48:in(Second Door)14:56:out(Second Door)
COMP123:John Sherrif15:04:in(Second Door)16:22:out(Second Door)
COMP123:John Sherrif16:34:in(Second Door)out

Understanding the Data: John Sherrif did Punch – IN at 08:33 for the first time and Punch – OUT at 09:04 for the first time. John did Punch – IN at 14:27 but forgot to do Punch – OUT. The ‘in’ signifies he/she forgot to do Punch IN and ‘out’ signifies vice versa.


  • Data Cleaning & Creating a table for status, Punch Code, and Emp Code.


import pandas as pd
# load data
df = pd.read_excel('bio.xlsx')
# removing NA values from the
# dataframe df
df = df.fillna("")
# removing all the blank rows
df1 = df.dropna(how='all')
# picking the rows where present
# or absent values are there from
# 14 no column
df1 = df1[df1['Unnamed: 14'].str.contains('sent')]
# Extracting only the Employee
# Names
df_name = df.dropna(how='all')
# from column no 3 we are picking
# Employee names
df_name = df_name[df_name['Unnamed: 3'].str.contains('Employee')]
# creating a new dataframe for Status,
# Punch Records and Employee Codes
zippedList = list(
    zip(df1['Unnamed: 14'], df1['Unnamed: 15'], df_name['Unnamed: 7']))
abc = pd.DataFrame(zippedList)





  • Extracting Data for Second Door only.



# Splitting the values by comma in 1
# no column (punch records)
for i in range(len(abc)):
    abc[1][i] = abc[1][i].split(",")
second_door = []
for i in range(len(abc)):
    s_d = []
    # Extracting all the values which contains
    # only :in(Second Door) or :out(Second Dorr)
    for j in range(len(abc[1][i])):
        if ':in(Second Door)' in abc[1][i][j]:
        if 'out(Second Door)' in abc[1][i][j]:


  • The punch record should start with ‘IN’ and end with ‘OUT’. Creating the pattern if it doesn’t follow.


# Punch Records should start with
# the keyword 'in'. If it doesn't
# follow then we wil add 'in' and it
# significants that the employee forgot
# to do punch in
in_time = []
for i in range(len(second_door)):
        if ':in(Second Door)' not in second_door[i][0]:
            second_door[i].insert(0, 'in')
# Punch Records should end with the keyword
# 'out'. If it doesn't follow then we wil
# add 'out' and it  significants that the
# employee forgot to do punch out
out_time = []
for i in range(len(second_door)):
        if ':out(Second Door)' not in second_door[i][(len(second_door[i]))-1]:
            second_door[i].insert(((len(second_door[i]))), 'out')


  • Creating the pattern ‘IN – OUT – IN – …..- OUT’. If someone forgot to do Punch – IN then we will put ‘IN’ & if someone forgot to do Punch – OUT then we will put ‘OUT’.


# final_in contains PUNCH - IN
# records for all employees
final_in = []
# final_out contains PUNCH - OUT
# records for all employees
final_out = []
for k in range(len(second_door)):
    in_gate = []
    out_gate = []
    # even position should be for Punch-
    # IN and odd position should be for
    # Punch - OUT if it doesn't follow
    # then we will create the pattern by
    # putting 'in' or 'out'
    for i in range(len(second_door[k])):
        if i % 2 == 0 and 'in' in second_door[k][i]:
                if 'out' not in second_door[k][i+1]:
        if i % 2 != 0 and 'out' in second_door[k][i]:
                if 'in' not in second_door[k][i+1]:
        if i % 2 != 0 and 'in' in second_door[k][i]:
                if 'out' not in second_door[k][i+1]:
        if i % 2 == 0 and 'out' in second_door[k][i]:
                if 'in' not in second_door[k][i+1]:
# final_in or final_out keep the
# records as a list under list form.
# to solve the problem we will merge the list
# aa contains merged list of Punch - IN
aa = final_in[0]
for i in range(len(final_in)-1):
    aa = aa + final_in[i+1]
# bb contains merged list of Punch - OUT
bb = final_out[0]
for i in range(len(final_out)-1):
    bb = bb + final_out[i+1]
for i in range(len(final_in[0])):
    print(final_in[0][i], '   ', final_out[0][i])


  • Creating the final table.


# Creating a dataframe called df_final
df_final = []
df_final = pd.DataFrame(df_final)
# Merging the Employee Names
Name = []
for i in range(len(abc)):
    for j in range(len(final_in[i])):
df_final['Name'] = Name
# Zipping the Employee Name, Punch -IN
# records and Punch - OUT records
zippedList2 = list(zip(df_final['Name'], aa, bb))
abc2 = pd.DataFrame(zippedList2)
# Renaming the dataframe
abc2.columns = ['Emp Code', 'Punch - IN', 'Punch - OUT']
abc2.to_excel('output.xlsx', index=False)
# Print the table


Hence, the raw biometric data has been structured and is converted to useful information.

 Attention geek! Strengthen your foundations with the Python Programming Foundation Course and learn the basics.  

To begin with, your interview preparations Enhance your Data Structures concepts with the Python DS Course. And to begin with your Machine Learning Journey, join the Machine Learning – Basic Level Course

My Personal Notes arrow_drop_up
Recommended Articles
Page :