Artificial Intelligence on Job-Hopping Forecasting: AI on Job-Hopping

Abstract

Artificial Intelligence (AI) technologies have been successfully applied to many fields, such as object detection and speech recognition. Among these applications, few consider cases where some feature values are missing or unreliable, such as in the prediction of job-hopping patterns where many profiles are incomplete, even though these missing features may be important for businesses (e.g. human resources and management). In this paper, we propose a novel AI technology, Sequential Optimization of Naive Bayesian (SONB), which not only makes predictions, but also learns the underlying pattern and automatically estimates missing or unreliable feature values. We analyzed several important job-hopping features and applied it to predict job-hopping patterns on many incomplete profiles. Our experiment shows SONB accurately estimates missing values and achieves state-of-the-art performance. In addition, the accuracy of deep learning is improved by 3% on the new dataset generated by SONB over the raw data. In summary, we introduce a novel AI technology for forecasting, which could also be used to estimate missing values in the input data. It is applied to a large (20,185,365 employee profiles) dataset and successfully predicts job-hopping patterns for employees based on their profiles, which could be a valuable resource for businesses.