CRISP-DM Methodology: GE Employee Attrition

CRISP-DM Methodology
Author
Affiliation
Course
Instructor
Due Date
Employee Attrition
Business Understanding
According to Smart Vision Europe (n.d.), the first step in the CRISP-DM process is to figure out what you want to accomplish from a business standpoint. Your company may have conflicting goals and constraints that must be appropriately balanced.
Background
GE has been losing many high-potential staff lately, and it is quite a concern to the organization. The cost of losing an employee accounts for 80 percent of the employee's annual salary. GE invests a lot of time and money in training new employees to remain competitive in the labor market. As a result, there is a need for a thorough and efficient model for detecting employees who are likely to churn. With the help of past records, the company can build a predictive model that can assess employees' profiles and classify them based on their likelihood to churn.
GE currently maintains employees' profiles on a web-based desktop system connected to an Oracle transactional database. The employee's database's metadata is also maintained in the same environment. However, the data information technology (IT) department has established a data warehouse. Every night, a predictive model would ingest the data and classify employees based on their likelihood of churn. An oracle transactional database would be used to store; data pipelines that link with the predictive model will be created. Key variables that would influence the employee’s likelihood to churn for distance from home, job level, monthly income, and job satisfaction would be further analyzed to examine their suitability in developing the employee attrition model.
Business problem: GE’s Human resource department would like to establish an employee attrition model that can predict and classify employees' profiles based on their likelihood to churn or not.
Research question: What predictive model can be developed and integrated into existing infrastructure to help GE’s human resource department predict employees’ likelihood of churn based on their profile?
Benefits
The solution to GE’s problem is to develop a machine learning model that will predict employees’ attrition based on their risk profile. A machine learning solution offers a wide range of benefits that include; (a) automation, (b) Better accuracy, and (c) Higher operation ability.
Automation: repetitive processes, for instance, screening each employee profile one after the other, can easily be replaced by innovative technologies. Data engineers can build data pipelines that batch employees' profiles and feed them into a predictive model for analysis. This process can happen in real-time or sequentially. Consequently, little human intervention is required resulting in cost savings.
Better Accuracy; Before a predictive model is rolled out into production in a real environment, machine learning models are evaluated on accuracy and other critical metrics that determine the robustness. In most cases, the threshold is customarily set above human abilities. For example, in our case, the model's accuracy should be above 90 percent accuracy compared to humans, which perform at around 85 perce...