Comparison of machine learning algorithms on data from the nuclear industry

Sample selection from a given dataset to validate machine learning models

“…This industrial problem aims at studying the fission products released in the primary circuit's water of the EDF nuclear reactors, during the load drop phase of the reactor cold shutdown. The available full dataset allows for N = 90 observations containing d = 25 covariates (describing the operation conditions of the reactor just before the shutdown) and the iodine activity level [12]. The goal is to model the event that this iodine activity level exceeds a specific threshold, that can have large impact on the scheduled planning, so on operational costs.…”