Employee attrition is a significant cost to an organization. A high attrition rate can lead to increased tangible costs such as training, recruitment, and on-boarding, as well as intangible costs such as project management and customer relationships. Attrition analysis and support in defining an optimal talent retention strategy. However, this requires a deep understanding of employee behavior. If HR directors designed talent retention strategies based on data and machine learning, companies could significantly reduce attrition and see an increase in employee productivity and company profitability. Together with attrition, the quantification of employee value and attrition cost is key to making optimal HR decisions.
A KNIME workflow collects employee data provided by the HR department and estimates the probability of each employee leaving the company. The dataset is then cleaned up for outliers, erroneous values, and/or representational structure. Employee attrition datasets are usually imbalanced for the attrition category; hence a rebalancing exercise is performed. Several models such as Random Forest, Logistic Regression, Naïve Bayes, and Gradient Boosted are trained, with the best performing model being chosen to score current employees. The workflow is deployed on KNIME Server and can be executed on demand or using the scheduling option, to update attrition probabilities. A dataset is generated with the model output, which the HR department uses for analysis on the profiles of the employees likely to leave.
With the native Tableau Integration in KNIME Analytics Platform, the results are exported directly to a Tableau dashboard. This provides business users with access to informative and useful visualizations quickly and easily. A Guided Analytics application, made accessible via the KNIME WebPortal, enables business users to make parametric changes to understand their impact on the attrition process. Business users can also simulate scenarios, which empowers them and promotes a data-driven decision-making culture.
Both workflows are available for download on KNIME Hub:
KNIME provides the ideal environment for building a classification machine learning model, and conducting an attrition analysis. It’s possible to build workflows using the appropriate connector for different data sources, transform data, and train machine learning models. KNIME Server enables these workflows to be executed at predefined times. The KNIME WebPortal allows business users to interact with these workflows by uploading their own data or updating the attrition probabilities on demand. The seamless integration with Tableau, enables users to create dashboards in order to dive deeper into the results.
This Innovation Note is available here as a PDF.