DETECTION OF TECHNICAL FAILURES ON PRODUCTION LINES USING MACHINE LEARNING, LINEAR AND BAYESIAN MODELS OF LOGISTIC REGRESSION

Abstract

In this work, we study the use of logistic regression in manufacturing failures detection. As a
data  set  for  the  analysis,  we  used  the  data  from  Kaggle  competition  “Bosch  Production  Line
Performance”.  We  considered  the  use  of  machine  learning,  linear  and  Bayesian  models.  For
machine  learning  approach,  we  analyzed  XGBoost  tree  based  classifier  to  obtain  high  scored
classification.  Using  the  generalized  linear  model  for  logistic  regression  makes  it  possible  to
analyze the influence of the factors under study. The Bayesian approach for logistic regression
gives  the  statistical  distribution  for  the  parameters  of  the  model.  It  can  be  useful  in  the
probabilistic analysis, e.g. risk assessment.  

Keywords: logistic regression; XGBoost; Bayesian inference; failure detection.