Missing value imputation method for cultural heritage sensing data based on the DSGAN-OD model
High-quality cultural heritage perception data is of great significance to cultural heritage conservation.However,due to the harsh natural environment conditions of cultural heritages,there are inevitably missing values in the data from sensing devices.And the missing data of the same category in cultural heritages have the characteristics of small sample.The existing missing values processing methods do not take into account the noise interference in cultural heritages data and the spatio-temporal correlation between small sample data,resulting in low accuracy of missing value interpolation.A missing value imputation model based on Semi-Supervised Generative Adversarial Networks(DSGAN-OD)is proposed.In this model,the multi-dimensional data are firstly de-noised and de-dimensional by DAE.Due to the unsupervised attribute of Generation Adversarial Networks,the classification label information in the cultural heritages data cannot be fully utilized.The low-dimensional expression vectors obtained by DAE were used as learning samples of semi-supervised generative adversarial networks(SemiGAN)to obtain features of missing datasets.Meanwhile,Order Decision(OD)method is used to determine the filling order of missing values according to the spatio-temporal correlation between data.Finally,the missing values are interpolated with the complete data generated by SemiGAN in this order.Experimental results on UCI standard dataset and cultural heritages temperature and humidity data show that compared with existing GAIN method,Random Forest method and MICE method,the accuracy of DSGAN-OD missing value interpolation model is improved by 21%,48.2%and 45.1%,respectively.
missing value imputationcultural heritages securityDSGAN-OD modelimputation order decision
国家科技期刊平台
NSTL
万方数据
