OPTIMIZING PARAMETERS OF THE TWO-LAYER PERCEPTRONS’ BOOSTING ENSEMBLE TRAINING FOR ACCURACY IMPROVEMENT IN WEAR STATE DISCONTINUOUS TRACKING MODEL REGARDING STATISTICAL DATA INACCURACIES AND SHIFTS

Abstract

There is a trial of optimization for improving accuracy in tracking metal tool wear states discontinuously, when the states’ finite set has been statistically tied to the set of representative wear influencing factors. Range of wear states is pre-sumed to be wholly sampled into those factors. The tracker is a static model based on boosting ensemble of two-layer percep-trons with nonlinear transfer functions. It successfully regards statistical data inaccuracies and shifts in a problem of tracking 24 wear states featured with 16 wear influencing factors. Having increased number of classifiers within the ensemble up to 30, the averaged gain with the optimized ensemble is about 56 % in respect of the best ensemble of three classifiers. Simi-larly, variance of tracking error rate over 24 wear states is about 53 % lower. Nearly the same results are registered when the ensemble is composed without training, but just setting every classifier’s weight to one thirtieth. To get the perfected accu-racy more, such equally-weighted compositions shall be investigated in the sequel.