Prediction Models for Professional Experience

Summary:

Salary prediction holds significant importance for both employers and employees. While prediction models can be trained using salary datasets, they typically lack information regarding a crucial factor for the salary: professional experience.

To improve salary predictions, Jannik Kiesel estimated the professional experience using regression models trained on the Socio-Economic Panel (SOEP) dataset. The regression models represent the culmination of his Bachelor’s thesis and are now available for download in .txt format for integration into regression models accessible via https://scikit-learn.org/.

Models:

• Linear Regression
• Quadratic Regression
• Cubic Regression

Please refer to the cited publication for more details on the estimation of professional experience.

In addition to the results in the paper, we further improved the performance (from an MAE of 3.74 years to 3.65 years) by using a cubic model that additionally takes into account the Occupational Area (the first digit of the 5-digit code of the occupation type according to the German Classification of Occupations 2010). We also improved our preprocessing by limiting professional experience to age minus 15 years (since 15 years is the earliest age at which one can start accumulating professional experience). We recommend that this logic also be used when applying the model. The improved model can be downloaded here: Cubic Regression with Occupational_Area

Citation:

Frank Eichinger, Jannik Kiesel, Matthias Dorner, and Stefan Arnold . 2023. Estimations of Professional Experience with Panel Data to Improve Salary Predictions. In Proceedings of the 43rd SGAI International Conference on Artificial Intelligence (AI-2023), Cambridge, UK. Springer.