Intelligent Classification Prediction Model For Multi-samples Data With Multi-dimensional Attributes And Its Application

Data sets including a lot of samples with multi-dimensional attributes are called as multi-samples data with multi-dimensional attributes.They generally exist in material science,medical care,aerospace,power electronics and other fields.By mining and analyzing these data and constructing their prediction models,the possible results of the specified target variables of multi-samples data with multi-dimensions in the above fields can be derived.However,although big data analysis and prediction theories have developed greatly in recent years,but it is difficult to obtain high prediction accuracy by directly using existing machine learning algorithms and statistical analysis methods to deal with multi-samples data with multi-dimensional attributes,due to their fuzziness,uncertainty,coupling and multi-dimensional attributes.The purpose of this research is that analysis and intelligent prediction modeling for multi-samples data with multidimensional attributes are proposed,and apply them to solve the classification and prediction problems for the curative effect of periodontitis,as well as to provide the effectiveness prediction models for multi-samples data with multi-dimensional attributes in other fields.The main innovative work of this paper includes:(1)On the basis of analyzing one-dimensional cloud,two-dimensional and multidimensional cloud modeling methods are proposed respectively in order to handle the multi-dimensional characteristics of big data.The algorithm process of two-dimensional and multi-dimensional precursor cloud droplets generation and the algorithm steps of two-dimensional and multi-dimensional single rule generator are given.The problem of low prediction accuracy caused by the fuzziness and uncertainty of multi-sample data with multi-dimensional is solved.(2)Based on the detailed analysis of the advantages and disadvantages of Xgboost,PCA and BP neural network,the Xgboost-PCA-BPNN combined prediction algorithm is proposed,which reduces the coupling of multi-sample data with multi-dimensional attributes and the effect of multidimensionality on prediction performance to improves the prediction accuracy of the obtained model.(3)On the basis of analyzing the 45,000 periodontitis' site data of an oral cavity hospital in Beijing,the proposed multi-dimensional cloud model and Xgboost-PCABPNN combination prediction algorithm were applied to predict the curative effect of periodontitis,and good prediction results were obtained.The test results show that the accuracy of Xgboost-PCA-BPNN combination prediction algorithm is 82%,which is better than multi-dimensional cloud model and other machine learning methods,such as logical regression,Xgboost and Xgboost-logic regression combination algorithm.