Analysing migrants’ fertility behaviour using machine learning techniques: an application of random survival forest to French data

Survival and event history analyses have become widely used techniques in life-course and longitudinal research. Machine learning methods such as survival trees and tree ensembles are a useful alternative to classical methods. This paper aims to illustrate the advantages of random survival forest (RSF). We apply the method to analyse migrant fertility: the probability of having a first, second and third birth among immigrants and their descendants in France. The results of the RSF indicate that even though immigrants have a higher probability of having a birth than natives, highly educated immigrants are much closer to natives in their childbearing patterns than low educated migrants. Our findings illustrate the usefulness of machine leaning techniques in two ways. First, RSF allows us to easily identify the most important predictors of a life event. Second, it allows us to detect and visualize interactions and therefore to identify groups of individuals with different survival probability.