In this project we'll work with exit surveys from the Technical and Further Education (TAFE) institute and the Department of Educationn, Training and Employment (DETE) in Queensland, Australia. All the information related to the DETE exit survey is here whereas the original TAFE exit survey is not available in this moment. So we'll work with a dataset with some slight modifications from the original information created by Dataquest.
The main aim of ther project'll be to clean and analyze the exit surveys in order to answer the following questions:
We'll start by importing the libraries and opening both datasets.
# Import libraries
import pandas as pd
import numpy as np
# Read datasets
dete_survey = pd.read_csv('C:/Users/User/Desktop/Dataquest/2.Introduction_to_Pandas_and_NumPy_for_Data_Analysis/datasets/dete_survey.csv')
tafe_survey = pd.read_csv('C:/Users/User/Desktop/Dataquest/2.Introduction_to_Pandas_and_NumPy_for_Data_Analysis/datasets/tafe_survey.csv')
# Dete survey
dete_survey.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 822 entries, 0 to 821 Data columns (total 56 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 ID 822 non-null int64 1 SeparationType 822 non-null object 2 Cease Date 822 non-null object 3 DETE Start Date 822 non-null object 4 Role Start Date 822 non-null object 5 Position 817 non-null object 6 Classification 455 non-null object 7 Region 822 non-null object 8 Business Unit 126 non-null object 9 Employment Status 817 non-null object 10 Career move to public sector 822 non-null bool 11 Career move to private sector 822 non-null bool 12 Interpersonal conflicts 822 non-null bool 13 Job dissatisfaction 822 non-null bool 14 Dissatisfaction with the department 822 non-null bool 15 Physical work environment 822 non-null bool 16 Lack of recognition 822 non-null bool 17 Lack of job security 822 non-null bool 18 Work location 822 non-null bool 19 Employment conditions 822 non-null bool 20 Maternity/family 822 non-null bool 21 Relocation 822 non-null bool 22 Study/Travel 822 non-null bool 23 Ill Health 822 non-null bool 24 Traumatic incident 822 non-null bool 25 Work life balance 822 non-null bool 26 Workload 822 non-null bool 27 None of the above 822 non-null bool 28 Professional Development 808 non-null object 29 Opportunities for promotion 735 non-null object 30 Staff morale 816 non-null object 31 Workplace issue 788 non-null object 32 Physical environment 817 non-null object 33 Worklife balance 815 non-null object 34 Stress and pressure support 810 non-null object 35 Performance of supervisor 813 non-null object 36 Peer support 812 non-null object 37 Initiative 813 non-null object 38 Skills 811 non-null object 39 Coach 767 non-null object 40 Career Aspirations 746 non-null object 41 Feedback 792 non-null object 42 Further PD 768 non-null object 43 Communication 814 non-null object 44 My say 812 non-null object 45 Information 816 non-null object 46 Kept informed 813 non-null object 47 Wellness programs 766 non-null object 48 Health & Safety 793 non-null object 49 Gender 798 non-null object 50 Age 811 non-null object 51 Aboriginal 16 non-null object 52 Torres Strait 3 non-null object 53 South Sea 7 non-null object 54 Disability 23 non-null object 55 NESB 32 non-null object dtypes: bool(18), int64(1), object(37) memory usage: 258.6+ KB
dete_survey.head()
ID | SeparationType | Cease Date | DETE Start Date | Role Start Date | Position | Classification | Region | Business Unit | Employment Status | ... | Kept informed | Wellness programs | Health & Safety | Gender | Age | Aboriginal | Torres Strait | South Sea | Disability | NESB | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 1 | Ill Health Retirement | 08/2012 | 1984 | 2004 | Public Servant | A01-A04 | Central Office | Corporate Strategy and Peformance | Permanent Full-time | ... | N | N | N | Male | 56-60 | NaN | NaN | NaN | NaN | Yes |
1 | 2 | Voluntary Early Retirement (VER) | 08/2012 | Not Stated | Not Stated | Public Servant | AO5-AO7 | Central Office | Corporate Strategy and Peformance | Permanent Full-time | ... | N | N | N | Male | 56-60 | NaN | NaN | NaN | NaN | NaN |
2 | 3 | Voluntary Early Retirement (VER) | 05/2012 | 2011 | 2011 | Schools Officer | NaN | Central Office | Education Queensland | Permanent Full-time | ... | N | N | N | Male | 61 or older | NaN | NaN | NaN | NaN | NaN |
3 | 4 | Resignation-Other reasons | 05/2012 | 2005 | 2006 | Teacher | Primary | Central Queensland | NaN | Permanent Full-time | ... | A | N | A | Female | 36-40 | NaN | NaN | NaN | NaN | NaN |
4 | 5 | Age Retirement | 05/2012 | 1970 | 1989 | Head of Curriculum/Head of Special Education | NaN | South East | NaN | Permanent Full-time | ... | N | A | M | Female | 61 or older | NaN | NaN | NaN | NaN | NaN |
5 rows × 56 columns
dete_survey['Role Start Date'].value_counts()
Not Stated 98 2012 65 2011 57 2010 46 2008 45 2013 41 2007 41 2009 38 2006 25 2004 22 2000 20 2005 20 2003 20 1999 19 1996 19 1989 18 1992 17 2002 16 2001 15 1998 15 1988 14 1995 13 1997 13 1975 12 1990 12 1986 11 1978 9 1976 9 1993 9 1991 8 1979 7 1994 7 1985 6 1981 5 1983 5 1987 5 1984 4 1982 4 1980 3 1970 2 1977 2 200 1 1974 1 1971 1 1973 1 1972 1 Name: Role Start Date, dtype: int64
# Tafe survey
tafe_survey.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 702 entries, 0 to 701 Data columns (total 72 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Record ID 702 non-null float64 1 Institute 702 non-null object 2 WorkArea 702 non-null object 3 CESSATION YEAR 695 non-null float64 4 Reason for ceasing employment 701 non-null object 5 Contributing Factors. Career Move - Public Sector 437 non-null object 6 Contributing Factors. Career Move - Private Sector 437 non-null object 7 Contributing Factors. Career Move - Self-employment 437 non-null object 8 Contributing Factors. Ill Health 437 non-null object 9 Contributing Factors. Maternity/Family 437 non-null object 10 Contributing Factors. Dissatisfaction 437 non-null object 11 Contributing Factors. Job Dissatisfaction 437 non-null object 12 Contributing Factors. Interpersonal Conflict 437 non-null object 13 Contributing Factors. Study 437 non-null object 14 Contributing Factors. Travel 437 non-null object 15 Contributing Factors. Other 437 non-null object 16 Contributing Factors. NONE 437 non-null object 17 Main Factor. Which of these was the main factor for leaving? 113 non-null object 18 InstituteViews. Topic:1. I feel the senior leadership had a clear vision and direction 608 non-null object 19 InstituteViews. Topic:2. I was given access to skills training to help me do my job better 613 non-null object 20 InstituteViews. Topic:3. I was given adequate opportunities for personal development 610 non-null object 21 InstituteViews. Topic:4. I was given adequate opportunities for promotion within %Institute]Q25LBL% 608 non-null object 22 InstituteViews. Topic:5. I felt the salary for the job was right for the responsibilities I had 615 non-null object 23 InstituteViews. Topic:6. The organisation recognised when staff did good work 607 non-null object 24 InstituteViews. Topic:7. Management was generally supportive of me 614 non-null object 25 InstituteViews. Topic:8. Management was generally supportive of my team 608 non-null object 26 InstituteViews. Topic:9. I was kept informed of the changes in the organisation which would affect me 610 non-null object 27 InstituteViews. Topic:10. Staff morale was positive within the Institute 602 non-null object 28 InstituteViews. Topic:11. If I had a workplace issue it was dealt with quickly 601 non-null object 29 InstituteViews. Topic:12. If I had a workplace issue it was dealt with efficiently 597 non-null object 30 InstituteViews. Topic:13. If I had a workplace issue it was dealt with discreetly 601 non-null object 31 WorkUnitViews. Topic:14. I was satisfied with the quality of the management and supervision within my work unit 609 non-null object 32 WorkUnitViews. Topic:15. I worked well with my colleagues 605 non-null object 33 WorkUnitViews. Topic:16. My job was challenging and interesting 607 non-null object 34 WorkUnitViews. Topic:17. I was encouraged to use my initiative in the course of my work 610 non-null object 35 WorkUnitViews. Topic:18. I had sufficient contact with other people in my job 613 non-null object 36 WorkUnitViews. Topic:19. I was given adequate support and co-operation by my peers to enable me to do my job 609 non-null object 37 WorkUnitViews. Topic:20. I was able to use the full range of my skills in my job 609 non-null object 38 WorkUnitViews. Topic:21. I was able to use the full range of my abilities in my job. ; Category:Level of Agreement; Question:YOUR VIEWS ABOUT YOUR WORK UNIT] 608 non-null object 39 WorkUnitViews. Topic:22. I was able to use the full range of my knowledge in my job 608 non-null object 40 WorkUnitViews. Topic:23. My job provided sufficient variety 611 non-null object 41 WorkUnitViews. Topic:24. I was able to cope with the level of stress and pressure in my job 610 non-null object 42 WorkUnitViews. Topic:25. My job allowed me to balance the demands of work and family to my satisfaction 611 non-null object 43 WorkUnitViews. Topic:26. My supervisor gave me adequate personal recognition and feedback on my performance 606 non-null object 44 WorkUnitViews. Topic:27. My working environment was satisfactory e.g. sufficient space, good lighting, suitable seating and working area 610 non-null object 45 WorkUnitViews. Topic:28. I was given the opportunity to mentor and coach others in order for me to pass on my skills and knowledge prior to my cessation date 609 non-null object 46 WorkUnitViews. Topic:29. There was adequate communication between staff in my unit 603 non-null object 47 WorkUnitViews. Topic:30. Staff morale was positive within my work unit 606 non-null object 48 Induction. Did you undertake Workplace Induction? 619 non-null object 49 InductionInfo. Topic:Did you undertake a Corporate Induction? 432 non-null object 50 InductionInfo. Topic:Did you undertake a Institute Induction? 483 non-null object 51 InductionInfo. Topic: Did you undertake Team Induction? 440 non-null object 52 InductionInfo. Face to Face Topic:Did you undertake a Corporate Induction; Category:How it was conducted? 555 non-null object 53 InductionInfo. On-line Topic:Did you undertake a Corporate Induction; Category:How it was conducted? 555 non-null object 54 InductionInfo. Induction Manual Topic:Did you undertake a Corporate Induction? 555 non-null object 55 InductionInfo. Face to Face Topic:Did you undertake a Institute Induction? 530 non-null object 56 InductionInfo. On-line Topic:Did you undertake a Institute Induction? 555 non-null object 57 InductionInfo. Induction Manual Topic:Did you undertake a Institute Induction? 553 non-null object 58 InductionInfo. Face to Face Topic: Did you undertake Team Induction; Category? 555 non-null object 59 InductionInfo. On-line Topic: Did you undertake Team Induction?process you undertook and how it was conducted.] 555 non-null object 60 InductionInfo. Induction Manual Topic: Did you undertake Team Induction? 555 non-null object 61 Workplace. Topic:Did you and your Manager develop a Performance and Professional Development Plan (PPDP)? 608 non-null object 62 Workplace. Topic:Does your workplace promote a work culture free from all forms of unlawful discrimination? 594 non-null object 63 Workplace. Topic:Does your workplace promote and practice the principles of employment equity? 587 non-null object 64 Workplace. Topic:Does your workplace value the diversity of its employees? 586 non-null object 65 Workplace. Topic:Would you recommend the Institute as an employer to others? 581 non-null object 66 Gender. What is your Gender? 596 non-null object 67 CurrentAge. Current Age 596 non-null object 68 Employment Type. Employment Type 596 non-null object 69 Classification. Classification 596 non-null object 70 LengthofServiceOverall. Overall Length of Service at Institute (in years) 596 non-null object 71 LengthofServiceCurrent. Length of Service at current workplace (in years) 596 non-null object dtypes: float64(2), object(70) memory usage: 395.0+ KB
tafe_survey.head()
Record ID | Institute | WorkArea | CESSATION YEAR | Reason for ceasing employment | Contributing Factors. Career Move - Public Sector | Contributing Factors. Career Move - Private Sector | Contributing Factors. Career Move - Self-employment | Contributing Factors. Ill Health | Contributing Factors. Maternity/Family | ... | Workplace. Topic:Does your workplace promote a work culture free from all forms of unlawful discrimination? | Workplace. Topic:Does your workplace promote and practice the principles of employment equity? | Workplace. Topic:Does your workplace value the diversity of its employees? | Workplace. Topic:Would you recommend the Institute as an employer to others? | Gender. What is your Gender? | CurrentAge. Current Age | Employment Type. Employment Type | Classification. Classification | LengthofServiceOverall. Overall Length of Service at Institute (in years) | LengthofServiceCurrent. Length of Service at current workplace (in years) | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 6.341330e+17 | Southern Queensland Institute of TAFE | Non-Delivery (corporate) | 2010.0 | Contract Expired | NaN | NaN | NaN | NaN | NaN | ... | Yes | Yes | Yes | Yes | Female | 26 30 | Temporary Full-time | Administration (AO) | 1-2 | 1-2 |
1 | 6.341337e+17 | Mount Isa Institute of TAFE | Non-Delivery (corporate) | 2010.0 | Retirement | - | - | - | - | - | ... | Yes | Yes | Yes | Yes | NaN | NaN | NaN | NaN | NaN | NaN |
2 | 6.341388e+17 | Mount Isa Institute of TAFE | Delivery (teaching) | 2010.0 | Retirement | - | - | - | - | - | ... | Yes | Yes | Yes | Yes | NaN | NaN | NaN | NaN | NaN | NaN |
3 | 6.341399e+17 | Mount Isa Institute of TAFE | Non-Delivery (corporate) | 2010.0 | Resignation | - | - | - | - | - | ... | Yes | Yes | Yes | Yes | NaN | NaN | NaN | NaN | NaN | NaN |
4 | 6.341466e+17 | Southern Queensland Institute of TAFE | Delivery (teaching) | 2010.0 | Resignation | - | Career Move - Private Sector | - | - | - | ... | Yes | Yes | Yes | Yes | Male | 41 45 | Permanent Full-time | Teacher (including LVT) | 3-4 | 3-4 |
5 rows × 72 columns
With the information showed above, we can make the following observations:
Role Start Date
inside dete_survey
with Not Stated
values. It means that they are missing, but they are not considered as Null
or Nan
.To start with the cleaning task, we'll again use pd.read_csv()
to specify that Not Stated
values should be represented as NaN
. Then, we'll drop the columns we are not interested in for our analysis.
# Not stated as NaN
dete_survey = pd.read_csv('C:/Users/User/Desktop/Dataquest/2.Introduction_to_Pandas_and_NumPy_for_Data_Analysis/datasets/dete_survey.csv', na_values='Not Stated')
print(dete_survey['Role Start Date'].head())
0 2004.0 1 NaN 2 2011.0 3 2006.0 4 1989.0 Name: Role Start Date, dtype: float64
# Drop useless columns in dete_survey
dete_survey_updated = dete_survey.drop(columns=dete_survey.columns[28:49], axis=1)
# Check the columns were dropped correctly
print(dete_survey_updated.columns)
len(dete_survey_updated.columns)
Index(['ID', 'SeparationType', 'Cease Date', 'DETE Start Date', 'Role Start Date', 'Position', 'Classification', 'Region', 'Business Unit', 'Employment Status', 'Career move to public sector', 'Career move to private sector', 'Interpersonal conflicts', 'Job dissatisfaction', 'Dissatisfaction with the department', 'Physical work environment', 'Lack of recognition', 'Lack of job security', 'Work location', 'Employment conditions', 'Maternity/family', 'Relocation', 'Study/Travel', 'Ill Health', 'Traumatic incident', 'Work life balance', 'Workload', 'None of the above', 'Gender', 'Age', 'Aboriginal', 'Torres Strait', 'South Sea', 'Disability', 'NESB'], dtype='object')
35
# Drop useless columns in tafe_survey
tafe_survey_updated = tafe_survey.drop(columns=tafe_survey.columns[17:66], axis=1)
# Check the columns were dropped correctly
print(tafe_survey_updated.columns)
len(tafe_survey_updated.columns)
Index(['Record ID', 'Institute', 'WorkArea', 'CESSATION YEAR', 'Reason for ceasing employment', 'Contributing Factors. Career Move - Public Sector ', 'Contributing Factors. Career Move - Private Sector ', 'Contributing Factors. Career Move - Self-employment', 'Contributing Factors. Ill Health', 'Contributing Factors. Maternity/Family', 'Contributing Factors. Dissatisfaction', 'Contributing Factors. Job Dissatisfaction', 'Contributing Factors. Interpersonal Conflict', 'Contributing Factors. Study', 'Contributing Factors. Travel', 'Contributing Factors. Other', 'Contributing Factors. NONE', 'Gender. What is your Gender?', 'CurrentAge. Current Age', 'Employment Type. Employment Type', 'Classification. Classification', 'LengthofServiceOverall. Overall Length of Service at Institute (in years)', 'LengthofServiceCurrent. Length of Service at current workplace (in years)'], dtype='object')
23
By dropping all the columns above, we have updated datasets which only have the information we are interested in to make our analysis. Thanks to this, we have a dataframe with 35 columns and another one with 23.
Now we'll standarize the columns names of both datasets, as we want to combine them in the same one.
# Rename dete_survey columns
dete_survey_updated.columns = dete_survey_updated.columns.str.lower().str.strip().str.replace(' ', '_')
# Check all names
print(dete_survey_updated.columns)
Index(['id', 'separationtype', 'cease_date', 'dete_start_date', 'role_start_date', 'position', 'classification', 'region', 'business_unit', 'employment_status', 'career_move_to_public_sector', 'career_move_to_private_sector', 'interpersonal_conflicts', 'job_dissatisfaction', 'dissatisfaction_with_the_department', 'physical_work_environment', 'lack_of_recognition', 'lack_of_job_security', 'work_location', 'employment_conditions', 'maternity/family', 'relocation', 'study/travel', 'ill_health', 'traumatic_incident', 'work_life_balance', 'workload', 'none_of_the_above', 'gender', 'age', 'aboriginal', 'torres_strait', 'south_sea', 'disability', 'nesb'], dtype='object')
# Rename tafe_survey columns
mapping = {'Record ID': 'id', 'CESSATION YEAR': 'cease_date', 'Reason for ceasing employment': 'separationtype', 'Gender. What is your Gender?': 'gender',
'CurrentAge.Current Age': 'age', 'Employment Type. Employment Type': 'employment_status', 'Classification. Classification': 'position',
'LengthofServiceOverall. Overall Length of Service at Institute (in years)': 'institute_service', 'LengthofServiceCurrent. Length of Service at current workplace (in years)': 'role_service'}
tafe_survey_updated = tafe_survey_updated.rename(mapping, axis=1)
# Check all names
print(tafe_survey_updated.columns)
Index(['id', 'Institute', 'WorkArea', 'cease_date', 'separationtype', 'Contributing Factors. Career Move - Public Sector ', 'Contributing Factors. Career Move - Private Sector ', 'Contributing Factors. Career Move - Self-employment', 'Contributing Factors. Ill Health', 'Contributing Factors. Maternity/Family', 'Contributing Factors. Dissatisfaction', 'Contributing Factors. Job Dissatisfaction', 'Contributing Factors. Interpersonal Conflict', 'Contributing Factors. Study', 'Contributing Factors. Travel', 'Contributing Factors. Other', 'Contributing Factors. NONE', 'gender', 'CurrentAge. Current Age', 'employment_status', 'position', 'institute_service', 'role_service'], dtype='object')
Next, let's remove more of the data we don't need. Focusing on the unique values in the separationtype
columns in each dataframe, we'll see that each contains a couple of different separation types. For this project, we are only interested in survey respondents who resigned, so their separation type contains the string 'Resignation'
.
# Values in dete_survey
dete_survey_updated['separationtype'].value_counts()
Age Retirement 285 Resignation-Other reasons 150 Resignation-Other employer 91 Resignation-Move overseas/interstate 70 Voluntary Early Retirement (VER) 67 Ill Health Retirement 61 Other 49 Contract Expired 34 Termination 15 Name: separationtype, dtype: int64
# Values in tafe_survey
tafe_survey_updated['separationtype'].value_counts()
Resignation 340 Contract Expired 127 Retrenchment/ Redundancy 104 Retirement 82 Transfer 25 Termination 23 Name: separationtype, dtype: int64
# Select only Resignation separationtype
dete_survey_updated['separationtype'] = dete_survey_updated['separationtype'].str.split('-').str[0]
# Check unique values
dete_survey_updated['separationtype'].value_counts()
Resignation 311 Age Retirement 285 Voluntary Early Retirement (VER) 67 Ill Health Retirement 61 Other 49 Contract Expired 34 Termination 15 Name: separationtype, dtype: int64
# Select Resignation separationtype for each dataframe
dete_resignations = dete_survey_updated[dete_survey_updated['separationtype']=='Resignation'].copy()
tafe_resignations = tafe_survey_updated[tafe_survey_updated['separationtype']=='Resignation'].copy()
In this step we'll verify that years in the cease_date
and dete_start_date
columns make sense. For that purpose:
cease_date
refers to the last year of the person's employment and the dete_start_date
is the first year of employment, all values above the current date wouldn't make sense.dete_start_date
was before 1940.# View unique values
dete_resignations['cease_date'].value_counts()
2012 126 2013 74 01/2014 22 12/2013 17 06/2013 14 09/2013 11 07/2013 9 11/2013 9 10/2013 6 08/2013 4 05/2012 2 05/2013 2 07/2012 1 2010 1 09/2010 1 07/2006 1 Name: cease_date, dtype: int64
# Extract year
dete_resignations['cease_date'] = dete_resignations['cease_date'].str.split('/').str[-1]
# Convert to float
dete_resignations['cease_date'] = dete_resignations['cease_date'].astype("float")
# Check values
dete_resignations['cease_date'].value_counts()
2013.0 146 2012.0 129 2014.0 22 2010.0 2 2006.0 1 Name: cease_date, dtype: int64
# Check values of dete_start_date in dete_resignations
dete_resignations['dete_start_date'].value_counts().sort_index(ascending=True)
1963.0 1 1971.0 1 1972.0 1 1973.0 1 1974.0 2 1975.0 1 1976.0 2 1977.0 1 1980.0 5 1982.0 1 1983.0 2 1984.0 1 1985.0 3 1986.0 3 1987.0 1 1988.0 4 1989.0 4 1990.0 5 1991.0 4 1992.0 6 1993.0 5 1994.0 6 1995.0 4 1996.0 6 1997.0 5 1998.0 6 1999.0 8 2000.0 9 2001.0 3 2002.0 6 2003.0 6 2004.0 14 2005.0 15 2006.0 13 2007.0 21 2008.0 22 2009.0 13 2010.0 17 2011.0 24 2012.0 21 2013.0 10 Name: dete_start_date, dtype: int64
# Check values of cease_date in tafe_resignations
tafe_resignations['cease_date'].value_counts().sort_index(ascending=True)
2009.0 2 2010.0 68 2011.0 116 2012.0 94 2013.0 55 Name: cease_date, dtype: int64
In the previous step we've found out that all values in both dataframes are inside the range defined, so we can consider all of them as correct values.
We can also see that the years in both dataframes don't completely align. For example, in tafe_resignations
there are some values referring to 2009, which are not in dete_resignations
. The tafe_resignations
dataframe also contains many more cease dates in 2010 than dete_resignations
. As we are not interested in analyzing the results by year, we'll leave the data the data like it is.
We have to recall that the question to answer in the project is: Are employees who have only worked for the institutes for a short period of time resigning due to some kind of dissatisfaction? What about employees who have been at the job longer?
The length of time an employee spent in a workplace is referred to as their years of service. This parameter is included in tafe_resignations
dataframe inside the column institute_service
, but there is no such column in dete_resignations
, so we should create it. To do so, we'll substract the dete_start_date
from cease_date
and create a new column with the results called institute_service
.
# New column with employee working years
dete_resignations['institute_service'] = dete_resignations['cease_date'] - dete_resignations['dete_start_date']
# Check the new column
dete_resignations['institute_service'].head()
3 7.0 5 18.0 8 3.0 9 15.0 11 3.0 Name: institute_service, dtype: float64
Now it's time to identify any employees who resigned due to dissatisfaction. Below are the columns used to categorize employees as "dissatisfied" from each dataframe:
Contributing Factors. Dissatisfaction
Contributing Factors. Job Dissatisfaction
job_dissatisfaction
dissatisfaction_with_the_department
physical_work_environment
lack_of_recognition
lack_of_job_security
work_location
employment_conditions
work_life_balance
workload
We'll create a column called dissatisfied
in each dataframe where we'll see if the employee indicated any of the factors above caused them to resign. After some changes, the columns dissatisfied
will have the following values:
True
: indicates a person resigned because they were dissatisfied in some way.False
: indicates the reason why a person resigned is not dissatisfaction.Nan
: indicates the value is missing.# Dissatisfaction in tafe_resignations
tafe_resignations['Contributing Factors. Dissatisfaction'].value_counts()
- 277 Contributing Factors. Dissatisfaction 55 Name: Contributing Factors. Dissatisfaction, dtype: int64
# Job Dissatisfaction in tafe_resignations
tafe_resignations['Contributing Factors. Job Dissatisfaction'].value_counts()
- 270 Job Dissatisfaction 62 Name: Contributing Factors. Job Dissatisfaction, dtype: int64
We'll create a function that enables to create the values we are interested in for the new column.
# Update values to True. False and NaN
def update_vals(val):
if pd.isnull(val)==True:
return np.nan
elif val=='-':
return False
else:
return True
# Apply the function
tafe_resignations['dissatisfied'] = tafe_resignations[['Contributing Factors. Dissatisfaction', 'Contributing Factors. Job Dissatisfaction']].applymap(update_vals).any(axis=1, skipna=False)
# Check the results
tafe_resignations['dissatisfied'].value_counts(dropna=False)
False 241 True 99 Name: dissatisfied, dtype: int64
# Dissatisfied column in dete_resignations
dete_resignations['dissatisfied'] = dete_resignations[['job_dissatisfaction', 'dissatisfaction_with_the_department', 'physical_work_environment', 'lack_of_recognition',
'lack_of_job_security', 'work_location', 'employment_conditions', 'work_life_balance', 'workload']].any(axis=1, skipna=False)
# Check the results
dete_resignations['dissatisfied'].value_counts(dropna=False)
False 162 True 149 Name: dissatisfied, dtype: int64
# Create a copy o both dataframes to avoid warning messages
tafe_resignations_up = tafe_resignations.copy()
dete_resignations_up = dete_resignations.copy()
Now, we're ready to combine the datasets. First of all, we'll add a column to each dataframe that will allow to distinguish between the two. Then we'll combine the dataframes and drop any remaining column we don't need for our analysis.
# New columns
dete_resignations_up['institute'] = 'DETE'
tafe_resignations_up['institute'] = 'TAFE'
# Combine datasets
combined = pd.concat([dete_resignations_up, tafe_resignations_up], ignore_index=True)
# Number of non null values in each column
combined.notnull().sum().sort_values()
torres_strait 0 south_sea 3 aboriginal 7 disability 8 nesb 9 business_unit 32 classification 161 region 265 role_start_date 271 dete_start_date 283 CurrentAge. Current Age 290 role_service 290 age 306 none_of_the_above 311 relocation 311 work_life_balance 311 traumatic_incident 311 ill_health 311 study/travel 311 maternity/family 311 workload 311 work_location 311 career_move_to_public_sector 311 employment_conditions 311 interpersonal_conflicts 311 job_dissatisfaction 311 career_move_to_private_sector 311 physical_work_environment 311 lack_of_recognition 311 lack_of_job_security 311 dissatisfaction_with_the_department 311 Contributing Factors. Interpersonal Conflict 332 Contributing Factors. Job Dissatisfaction 332 Contributing Factors. Dissatisfaction 332 Contributing Factors. Maternity/Family 332 Contributing Factors. Ill Health 332 Contributing Factors. Career Move - Self-employment 332 Contributing Factors. Career Move - Private Sector 332 Contributing Factors. Career Move - Public Sector 332 Contributing Factors. Travel 332 Contributing Factors. Other 332 Contributing Factors. Study 332 Contributing Factors. NONE 332 Institute 340 WorkArea 340 institute_service 563 gender 592 employment_status 597 position 598 cease_date 635 institute 651 separationtype 651 dissatisfied 651 id 651 dtype: int64
# Drop any columns with less than 500 non null values
combined_updated = combined.dropna(axis=1, thresh=500).copy()
# Info of the new dataframe
combined_updated.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 651 entries, 0 to 650 Data columns (total 9 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 id 651 non-null float64 1 separationtype 651 non-null object 2 cease_date 635 non-null float64 3 position 598 non-null object 4 employment_status 597 non-null object 5 gender 592 non-null object 6 institute_service 563 non-null object 7 dissatisfied 651 non-null bool 8 institute 651 non-null object dtypes: bool(1), float64(2), object(6) memory usage: 41.4+ KB
After the changes above, we have a dataframe which combines the information of the previous dataframes and only contains the columns we are interested in.
Before making the final analylisis, we'll have to clean up the institute_service
column because it currently contains values in a couple different forms.
To analyze the data, we'll convert these numbers into categories. We'll base our analysis on this article, which makes the argument that understanding employee's needs according to career stage instead of age is more effective.
We'll use the following categories:
# Check unique values in institute_service
combined_updated['institute_service'].value_counts()
Less than 1 year 73 1-2 64 3-4 63 5-6 33 11-20 26 5.0 23 1.0 22 7-10 21 3.0 20 0.0 20 6.0 17 4.0 16 9.0 14 2.0 14 7.0 13 More than 20 years 10 13.0 8 8.0 8 15.0 7 20.0 7 10.0 6 14.0 6 12.0 6 17.0 6 22.0 6 18.0 5 16.0 5 11.0 4 23.0 4 24.0 4 32.0 3 39.0 3 19.0 3 21.0 3 36.0 2 25.0 2 30.0 2 26.0 2 28.0 2 49.0 1 41.0 1 27.0 1 42.0 1 29.0 1 34.0 1 31.0 1 33.0 1 35.0 1 38.0 1 Name: institute_service, dtype: int64
# Extract years of service from institute_service
combined_updated['institute_service'] = combined_updated['institute_service'].astype('str').str.extract(r'(\d+)')
# Change to float type
combined_updated['institute_service'] = combined_updated['institute_service'].astype('float')
# Check type and unique values of the column
combined_updated['institute_service'].value_counts()
1.0 159 3.0 83 5.0 56 7.0 34 11.0 30 0.0 20 20.0 17 6.0 17 4.0 16 9.0 14 2.0 14 13.0 8 8.0 8 15.0 7 17.0 6 10.0 6 22.0 6 14.0 6 12.0 6 18.0 5 16.0 5 24.0 4 23.0 4 21.0 3 19.0 3 39.0 3 32.0 3 25.0 2 28.0 2 26.0 2 36.0 2 30.0 2 34.0 1 27.0 1 29.0 1 42.0 1 33.0 1 41.0 1 35.0 1 49.0 1 38.0 1 31.0 1 Name: institute_service, dtype: int64
# Function that maps each year to one of the career stage
def career_stages(val):
if val < 3:
return 'New'
elif 3 <= val < 7:
return 'Experienced'
elif 7 <= val < 11:
return 'Established'
elif pd.isnull(val):
return np.nan
else:
return 'Veteran'
# Apply the function and create a new column
combined_updated['service_cat'] = combined_updated['institute_service'].apply(career_stages)
#Check the result
combined_updated['service_cat'].value_counts()
New 193 Experienced 172 Veteran 136 Established 62 Name: service_cat, dtype: int64
Now we'll replace the NaN
values in dissatisfied
column with the most common value there(False
). Then, we'll calculate the percentage of employees who resigned due to dissatisfaction in each service_cat group
and plot the results to make the analysis more visual.
# Unique values in dissatisfied column
combined_updated['dissatisfied'].value_counts(dropna=False)
False 403 True 248 Name: dissatisfied, dtype: int64
# Replace NaN with the most common values
combined_updated['dissatisfied'] = combined_updated['dissatisfied'].fillna(False)
# Check the result
combined_updated['dissatisfied'].value_counts()
False 403 True 248 Name: dissatisfied, dtype: int64
# Percentage of dissatisfied employees in each service category
dissatisfied_percentage = combined_updated.pivot_table(values='dissatisfied', index='service_cat')
# Plot the results
%matplotlib inline
dissatisfied_percentage.plot(kind='bar')
<AxesSubplot:xlabel='service_cat'>
It's clear that employees with more than 6 years of service are more likely to resign the job due to dissatisfaction. However, we need to handle the rest of missing values to complete the whole analysis.