Tracking Data and Predicting Employee Burnout
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Background:
Data shows a large issue a lot of companies are facing is employee burnout. As employees burn out they become much less productive. It is estimated that employee burnout costs companies between $125 billion and $195 billion each year. By preventing burnout a company can save money and increase the morale of their employees. This project's goal is to look at what factors may influence employee burnout and possibly to predict it.
The Data:
The data was derived from a data set from Kaggle. The data set contains responses of a from employees across different industries. Some of the key data points are: date of joining, gender, company type (service or product), work from home availability, resource allocation (amount of resources and employee is allocated to work such as working hours), employee tenure designation, and finally a mental fatigue score. This project looks at which one of these factors determine the mental fatigue score.
Total spread of Mental Fatigue Score
Predictive Data:
When analyzing the data I decided to go by P-value and AIC to determine what data would be most predictive when analyzing employee burnout. The first piece of data used with a p-value of <.001 is work from home availability.
The second piece of data with a p-value of <.001 was resource allocation.
And finally the third piece of data with a p-value that ranges by factor is tenure designation.
Multiple Linear Regression Analysis:
In order to for the model to remain simple to the burn out fatigue score was not transformed. When attempting to transform the data the multiple R squared was mostly unaffected. In total, the model had a multiple R squared of 66%. This shows that employee burnout can be somewhat predicted. Not having the ability to work from home, having more work to do, and more responsibility will increase an employee's mental fatigue.
Coefficients of Model
Recommendation and Testing:
The recommendations made for companies attempting to lower mental fatigue score are as follows:
- Strongly consider work from home options
- Consider hiring more employees in roles with high resource allocation
- Promote work/life balance amongst senior employees
Success can be determined by resurveying the employees in 6 months to test if mental fatigue score is lowered.
Future Work:
In the future to help improve the model I would like to look at survey results from one specific company this way it will make the dependent variables more standard. For example an employee designation of 3 in company X may look different in company Y. I believe this will make the model more accurate. However, if the survey were to remain broad it may be beneficial to notate the industry the employee works in. This will also help standardize the data as well as give a better idea as to how the date affects mental fatigue (for example an accountant will probably answer differently in February vs November).
About Author
