Strategisk Personalplanering : Förutsäga Personalomsättning och Identifiera Drivkrafter Bakom Avgångar med Maskininlärning

In an increasingly competitive labour market, workforce planning strategies is critical to maintaining organisational stability, knowledge continuity and a sustainable workforce. For large multinational companies like Ericsson this is especially important, but it also presents certain challenges. One such challenge is proactively managing fluctuations in employee turnover and ensuring that the right competencies are in place at the right time, a key factor in sustaining long-term productivity and competitiveness. This thesis investigates the application of supervised machine learning algorithms to support organisations’ strategic workforce planning by predicting employee attrition. The aim is to provide general insights that can strengthen talent strategies and organisational planning across various industries. Ericsson is used as a case study to illustrate how these methods can be applied in practice within a large multinational corporation. Using historical employee data several machine learning models were evaluated: XGBoost, Random Forest, Logistic regression, K-Nearest Neighbours and Gaussian Naive Bayes, to determine which model most accurately predicts attrition. The results show that XGBoost consistently performs better, achieving the highest scores on several evaluation metrics, including accuracy, precision, recall, F1 score and AUC. The study also identifies the most critical factors for predicting attrition, such as reporting month, employee/manager impact assessments, and characteristics related to organisational structure and job level, which offer valuable insights into which variables are most indicative of an increased likelihood that an employee will leave or not. Analysis of these factors reveals valuable patterns, including elevated attrition rates in early career stages (Job Stage 01) and a slight increase in later stages, highlighting a challenge connected to Ericsson’s broader workforce strategy concerning age distribution. By identifying these patterns and variables, the study provides data-driven insights that can strengthen an organisations’ workforce planning and support a resilient organisational structure.