Machine Learning Approach for Employee Attrition Analysis

International Journal of Trend in Scientific Research and Development (IJTSRD)
Conference Issue | March 2019 Available Online: www.ijtsrd.com e-ISSN: 2456 - 6470
Fostering Innovation, Integration and Inclusion Through
Interdisciplinary Practices in Management
@ IJTSRD | Unique Paper ID - IJTSRD23065 | Conference Issue | FIIITIPM - 2019 | March 2019 Page: 62
Machine Learning Approach for Employee Attrition Analysis
Dr. R. S. Kamath1, Dr. S. S. Jamsandekar2, Dr. P. G. Naik3
1Associate Professor, 2Assistant Professor, 3Professor
1,2,3Department of Computer Studies,
1,2,3Chhatrapati Shahu Institute of Business Education and Research, Kolhapur, Maharashtra, India
Organised By:
Management Department, Chhatrapati
Shahu Institute of Business Education
and Research, Kolhapur, Maharashtra
How to cite this paper: Dr. R. S. Kamath
| Dr. S. S. Jamsandekar | Dr. P.G. Naik
"Machine Learning Approach for
Employee AttritionAnalysis"Published in
International Journal of Trend in
Scientific Research and Development
(ijtsrd), ISSN: 2456-6470, Special Issue |
Fostering Innovation, Integration and
Inclusion Through
Interdisciplinary
Practices in
Management,
March 2019, pp.62-
67, URL:
https://www.ijtsrd.
com/papers/ijtsrd
23065.pdf
ABSTRACT
Talent management involves a lot of managerial decisions to allocate right
people with the right skills employed at appropriate location and time. Authors
report machine learning solution for Human Resource (HR) attrition analysis
and forecast. The data for this investigation is retrieved from Kaggle, a Data
Science and Machine Learning platform [1]. Present study exhibits performance
estimation of various classification algorithms and compares the classification
accuracy. The performance of the model is evaluated in terms of Error Matrix
and Pseudo R Square estimate of error rate. Performance accuracyrevealedthat
Random Forest model can be effectively used for classification. This analysis
concludes that employeeattrition depends moreonemployees’satisfaction level
as compared to other attributes.
INTRODUCTION
The process to identifying the existing talent in an organization is among the top
talent management challenges and the important issue. For every organization,
human resource plays a vital role in all strategic decisions. Satisfied, highly-
motivated and loyal employees represent the basis of a company and which in
turn have impacts on the productivity of an organization.
The prime objective of the present study is to analyze why some of the best and
most experienced employees are leaving prematurely. This analysis also wishes
to predict which valuable employees will leave next.
The rest of paper is designed as follows; Introduction followed by the materials
and methods utilized in the present study. Then the third section summarizesthe
results and discussions of the HR attrition analysis. The conclusion at the end
justifies the suitability of Random Forest model for this talent mining.
Materials and Methods
The dataset for the present analysis is taken from Kaggle, Machine Learning platform [1]. This is the simulated dataset
comprising 15000 employee records classified into two categories (left or not left) based on satisfaction level, latest
evaluation, number of project worked on, average monthlyhours, timespendin thecompany,workaccident,promotion within
the past 5 years, department and salary. Table 1 gives description of employee dataset.
Table 1: Employee dataset description for talent mining
Attribute Description Data Type
satisfaction_level Level of satisfaction (0-1) Numeric
last_evaluation Time since last performance evaluation (in Years) Numeric
number_project Number of projects completed while at work Numeric
average_montly_hours Average monthly hours at workplace Numeric
time_spend_company Number of years spent in the company Numeric
Work_accident Whether the employee had a workplace accident Numeric
Left Whether the employee left the workplace or not (1 or 0) Numeric
promotion_last_5years Whether the employee was promoted in the last five years Numeric
sales Department in which they work for String
salary Relative level of salary (high) String
IJTSRD23065

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
This section explores details of experiment conducted for employee attrition analysis and forecasting. The present study is
carried out using R and Rattle data mining platform [4]. Figure 1 shows summary of the HR dataset. Dataset is partitioned
randomlyinto training, testing and validation with division 70%, 15 % and 15% respectively. We used the trainingdatasetfor
parameter adjustment of model whereas validation set to control learning process.
Figure 1: Dataset exploration – Summary
Among the vast machine learning algorithms, authors have picked Decision Tree, Random Forest, Support Vector Machine
(SVM), and Linear Regression techniques to build the model. These algorithms are based on supervised learning and best
known for building prediction models [8]. Supervised learning algorithms try to model relationships and dependencies
between the target prediction output and the input features/ predictors such that we can predict the output values for new
data based on those relationships which it learned from the previous data sets.
Figure 2 explains Decision tree modeling of HR data. It begins with a root node “satisfaction level”, that part into different
branches, prompting to further nodes, each of which may additionally part or else end as a leaf node. Connected with each
nonleaf node will be a test or question that figures out which branch to take after [7]. The leaf nodes indicatetheattrition sates
whether the employee “left” or “not left”. Figure 3 gives pictorial representation of Decision tree thus derived.

Figure 2: Decision tree modeling
Figure 3: Decision tree for HR attrition status
Figure 4 explains Random Forest Modeling for HR attrition analysis. RANDOMFOREST packagein Renvironmentis employed
here to analyze model structure[5-6]. RF builds manydecision trees usingrandom subset of data andvariables.Rattle provides
access to three parameters such as the number of trees, sample size and number of variables for tuning the models.

Figure 4: Summary of the Random Forest Model
Figure 5 explains Support Vector Machine (SVM) designedfortheattrition analysisof employeedata. SVM searches forsupport
a vector that separates the class.
Figure 5: Summary of SVM Model
Figure 6 explains Linear Regression Model. It is the traditional method for fitting a statistical model to data. It is
appropriate since the target variable “attrition status” is numeric.
Figure 6: Summary of Logistic Regression model
Results and Discussion
The present investigation employed different prediction algorithms to analyze employee attrition status and likelihood of
retention-attrition of employees. The performance of the model is evaluated in terms of Error Matrix and Pseudo R Square
estimate of error rate. An error matrix shows the true outcomes against the predicted outcomes. It is also known as confusion
matrix. Table 2 explains performance analysis of these classifiers in terms of error matrix.

Table 2: Performance Analysis of the Classifiers
Model Error Matrix
Decision Tree
Random Forest
Support Vector Machine
Liner Model
Figure 7, the “Predicted versus Observed” plot shows the performance analysis of all the four models. The plot displays the
predicted values against the observed values. The Pseudo R-Squared, square of the correlation between the predicted and
observed values. The closer to 1, is the acceptable one. Table 3 gives Pseudo R-Square values for these four models.
Figure 7: “Predicted versus Observed plot” for classifiers

Table 1: Performance accuracy of classifiers
Classifier Pseudo R-square
Decision Tree 0.8473
Random Forest 0.9773
Support Vector Machine 0.8315
Linear Regression 0.2299
Confusion matrix and “Predicted versus Observed” plot concludes that Random Forest is the appropriatemodelforanalysis of
Employee attrition as compared to the other algorithms considered in this studyandtheunderlined data.Figure8 explains the
relative importance of HR dataset attributes using Gini importanceandPermutation importancemeasures. Basedon thesetwo
measures, it reveals that employees’ “satisfaction level” is the predominant predictor of employee attrition.
Figure 8: Dependency of employee attrition status on other attributes
Conclusion
Authors have explored a machine learning solution for HR
attrition analysis and forecast. Present study exhibits
performance estimation of various classification algorithms
and compares the classification accuracy. The performance
of the model is evaluated in terms of Error Matrix and
Pseudo R Square estimate of error rate. Performance
accuracy revealed that Random Forest model can be
effectively used for classification. The result also concludes
that employee attrition depends more on employees’
satisfaction level as compared to other attributes.
References:
[1] Retrieved on 30th Dec, 2017 from
https://www.kaggle.com/ludobenistant/hr-analytics-
1/data
[2] Boudreau, J. W. – Ramstad, P. M.: Beyond HR. Boston.
Harvard Business School Press, 2007.ISBN978-1-4221-
0415-6.
[3] https://www.infogix.com/blog/machine-learning-vs-
statistical-modeling-the-real-difference, accessed date
28/012/2017
[4] Graham, W. Data Mining with Rattle and R: The Art of
Excavating Data for KnowledgeDiscovery,Springer,DOI
10.1007/978-1-4419-9890-3
[5] Breiman, L. (2005), Random Forest. Machine
Learning,45, 5-32
[6] Andy, L., & Matthew, W. (2002). Classification and
Regression by random Forest, R News, 2(3)
[7] R. S. Kamath, R. K. Kamat (2016), Modeling of Random
Textured Tandem Silicon Solar Cells Characteristics:
Decision Tree Approach, Journal of Nano and Electronic
Physics, Vol. 8 No 4(1), 04021(4pp)
[8] R. S. Kamath, R .K. Kamat (2016), Supervised Learning
Model for Kick starter Campaigns with R Mining,
International Journal of Information Technology,
Modeling and Computing (IJITMC), Vol. 4, No.1,
February, 19-30
Copyright © 2019 by author(s) and
International Journal of Trend in
Scientific Research and Development
Journal. This is an Open Access article distributed under the
terms of the Creative Commons Attribution License (CC BY
4.0) (http://creativecommons.org/licenses/by/4.0)

Machine Learning Approach for Employee Attrition Analysis

More Related Content

What's hot (20)

Similar to Machine Learning Approach for Employee Attrition Analysis (20)

More from ijtsrd (20)

Recently uploaded (20)

Machine Learning Approach for Employee Attrition Analysis