loader
banner

Boost Attrition Analysis with Machine Learning

Share on facebook
Share on twitter
Share on linkedin
Share on whatsapp
Share on pinterest

Share

Share on facebook
Share on twitter
Share on linkedin
Share on whatsapp
Share on pinterest

As a working professional we all know how the system works in all fields of business. Human Resource is the only asset which has its various factors affecting working and productiveness. Employees are considered as the backbone of any organization and here Attrition Analysis Comes into a Picture.

Their contribution and skills define the failure or success of an entity. It is a complex job for an organization to hire a talent, once hired the training part, all the experience, workflow understanding count with that employee as he becomes accustomed with the environment.

Unlike the machines, they cannot be repaired if damaged but can be replaced. Replacing a machine is much easier than replacing an employee.

The Problem

The organization has to face many problems if an experienced employee leaves the company for better future opportunities. There can be many reasons for dissatisfaction among employees for which they decide to change the company. 

Employee Attrition is a costly problem for any company as the cost of replacement is quite high. This demands proper analysis and solutions to counter this problem which can be a hectic job because there are many factors which contribute to satisfaction of an employee involving economic, environmental and emotional factors. 

Some of the reasons for an employee leaving an organization could be better paying job outside, poor relationship with manager or colleagues, job dissatisfaction, salary not expected, lack of opportunity for career development, workload, overtime etc.

Regarding attrition there can be other thing the manager wants to know like What are key indicators of affecting employee attrition, what can be the best strategy should be adopted for a particular employee for retaining them.

The Solution

In order to tackle this problem, this demands a system that uses the historical data of employees to analyze reasons for employee attrition. This system predicts who employees will have a high chance to leave the organization so that various corrective actions can be taken to prevent attrition and to ensure that employees stay in the organization.

Also, it provides how predicted attrition distributed among various features affecting attrition. Some of the employee retention strategies that can be opted for are motivating employees, giving new opportunities, providing a cooperative environment, resolving issues, new roles, taking constant feedback from employees, etc. 

There are different machine learning algorithms which help in building optimal systems for this issue like Random Forest, Logistic Regression, SVM (Support Vector Machine), KNN (K-Nearest Neighbor), Decision Tree. Also, analysis level can be enhanced with graphical representation for taking proper measure to build the model. 

The Data

To build the system, a sample dataset has been used to test out various ML algorithms. The dataset includes 1470 employees and 35 features having factors affecting attrition and attrition with yes and no status. “Yes” if an employee leaves the organization and “No” if not.

The data have various features like ‘Age’, ‘Attrition’, ‘BusinessTravel’, ‘DailyRate’, ‘Department’,  ‘DistanceFromHome’, ‘Education’, ‘EducationField’, ‘EmployeeCount’, ‘EmployeeNumber’, ‘EnvironmentSatisfaction’, ‘Gender’, ‘HourlyRate’, ‘JobInvolvement’, ‘JobLevel’, ‘JobRole’, ‘JobSatisfaction’, ‘MaritalStatus’, ‘MonthlyIncome’, ‘MonthlyRate’, ‘NumCompaniesWorked’,  ‘OverTime’, ‘PercentSalaryHike’, ‘PerformanceRating’, ‘RelationshipSatisfaction’, ‘StandardHours’, ‘StockOptionLevel’, ‘TotalWorkingYears’, ‘TrainingTimesLastYear’, ‘WorkLifeBalance’, ‘YearsAtCompany’, ‘YearsInCurrentRole’, ‘YearsSinceLastPromotion’, ‘YearsWithCurrManager’. It’s very prominent to do analysis of data before jumping to the modelling part. Let’s see how data flow goes. 

Given image showing how each feature is correlated to each other. Correlation measures the extent to which those two variables are related. Correlation defined by scales with given range. Correlation helps in choosing features for model that which feature contributing more in affecting attrition.

The Pearson correlation shows there is strong correlation between some features like job level and monthly income, job level and total working hours and other attributes. Targeted attributes have very less strong correlation with other attributes. Attrition is more likely to depend on multiple attributes rather than a single one.

Attrition Analysis

The System

To build the effective and optimal model testing and analyzing performance on various algorithms in python language, out of which Random Forest outperforms well with 84% accuracy. Accuracy was one of the main criteria in choosing the best algorithm.

The result will suggest which employees have chances to leave the company or not with probability percentages. A proper analysis with visual representation is shown below. These visuals are generated with the resulting output as how the predicted attrition status is distributed among various attributes.

Attrition Analysis

Above analysis was done among top attributes with a high rate of importance affecting attrition status. The conclusion from the above analysis shows, employees under age of 25 , single, male , travel rarely , do overtime and less experience in current company would leave mostly. 

The conclusion

HR analytics especially attrition prediction is up and hot area that many companies with highly data driven and strong analytical skills can improve efficiency in employee management. Understanding why and when are most likely to leave can lead to various benefits:

  • Planning new hiring in advance
  • Improving employee retention rate
  • Boosting up employee morale
  • Increasing employee satisfaction

Do You have Something In mind. Let me Know in Comment. I read every comment and more than happy to reply you back.

Like this article? Share this post

Share on facebook
Share on twitter
Share on linkedin
Share on pinterest
Share on print
Share on email

Leave a Reply

Your email address will not be published. Required fields are marked *

Recent Posts

Subscribe for News Letter:

Subscribe to Our Monthly Newsletter