An improved attribute recognition algorithm of overhead line engineering evaluation based on confidence dispersion

Abstract

Attribute recognition algorithms are widely used in engineering evaluation. However, the confidence parameters in this algorithm are usually chosen empirically, which has a great influence on the accuracy of engineering evaluation. In this paper, in order to improve the accuracy of the algorithm, a so-called confidence dispersion technical parameter is proposed to describe the influences of the sample data on confidence parameters in attribute recognition model. The numerical dispersion characteristics and the statistical distribution of the optimal confidence intervals are analyzed and the validity of confidence dispersion index has been proved by empirical model derivation and experimental simulation. Then a data-driven attribute recognition evaluation method is proposed based on the proposed confidence dispersion technical parameter, with the confidence parameters adaptive to sample data. According to contrastive simulation results on the technical design evaluation database of 220kV overhead line engineering projects, it has been verified that the proposed algorithm is feasible and effective, which also provides a new idea for the future design quality evaluation tasks of over-head line engineering projects.