Predictive monitoring using machine learning algorithms and a real‐life example on schizophrenia

Abstract

Predictive process monitoring aims to produce early warnings of unwanted events. We consider the use of the machine learning method extreme gradient boosting as the forecasting model in predictive monitoring. A tuning algorithm is proposed as the signaling method to produce a required false alarm rate. We demonstrate the procedure using a unique data set on mental health in the Netherlands. The goal of this application is to support healthcare workers in identifying the risk of a mental health crisis in people diagnosed with schizophrenia. The procedure we outline offers promising results and a novel approach to predictive monitoring.

Machine learning integration of multimodal data identifies key features of blood pressure regulation

“…The XGBoost algorithm has been successfully used for a wide range of medical applications, including disease diagnosis, survival estimation, outcomes, prognosis, drug research and development. 38 …”

“…The XGBoost algorithm has been successfully used for a wide range of medical applications, including disease diagnosis, survival estimation, outcomes, prognosis, drug research and development. 38 XGBoost was created by Chen and Guestrin 39 as an ensemble of multiple decision trees. Decision trees are a robust ML model capable of a high degree of accuracy and interpretability.…”

Comparison of threshold tuning methods for predictive monitoring

“…As mentioned in the main text, the original algorithm proposed by Huberts et al 13 is altered to accommodate our particular dataset characteristics and for comparison purposes. It should be noted that these alterations may not be suitable for other situations; in their study, the algorithm in its original set-up showed good performance.…”

“…After obtaining the threshold, a new model is fitted on the entire training set for prediction on the test set. The set-up is based on the work by Huberts et al 13 ; for changes made to the original algorithm, we refer to Appendix B. Let 𝐹𝐴𝑅 𝑣𝑎𝑙 be defined as the FAR obtained on the internal validation set.…”

“…For instance, Huberts et al 13 applied logistic hierarchical regression and XGBoost in a mental health setting and proposed a data-driven tuning procedure to achieve a desired FAR. Being based on a split-sample approach, this method however comes with a drawback: the model needs to be re-trained several times, and therefore this technique comes with a high computational cost.…”