Research On Uncertain SVM Improvement Model And Its Application In Pellet

A large number of uncertain information data exist in the objective world.If the training set contains uncertain information,the SVM is already weak.The development of blast furnace concentrates and environmental protection requirements have increased,and pellets have become an important choice for blast furnace feeding.Performance indicators such as the compressive strength of pellets and the bonding rate are important conditions for the stable production of blast furnaces,but there is uncertainty information in these factors.Therefore,the focus is on an uncertainty SVM improved algorithm model,which is in the prediction of the compressive strength and bonding rate of the pellets.The main work and research results are as follows.First,for samples of uncertain information,starting from the perspective of information cognition,the key to the classification of uncertainty cognition.Analyze the sources of uncertain information,and introduce mathematical knowledge related to fuzzy sets.The fuzzy membership processing method of uncertain information samples is proposed to effectively eliminate or reduce the influence of uncertain information samples on input samples.Second,according to the fuzzy set theory knowledge,combined with fuzzy set distance measurement and fuzzy set similarity construction principle,select the appropriate fuzzy membership function.The cut-set method is used to optimize the fuzzy membership value of the sample to avoid the problem of complex membership calculation,and the reliability of the model can be analyzed.The sample data processed by the fuzzy membership degree is input into SVM training and learning,and an improved SVM model with uncertainty is designed.The cross-validation method of grid search is used to optimize the parameters of the model core parameters c and g.Based on the principle of error analysis,the performance of the improved model is evaluated,and the design of the target algorithm(uncertainty SVM improved algorithm)is realized.Third,it is based on the improved uncertainty SVM.Considering the uncertainties of the compressive strength and bonding rate of the pellets and the uncertainty of the factors affecting these performance indicators,150 sets of experimental data were selected to verify the improved algorithm during pellet production.The root mean square error,average absolute error and average absolute percentage error are used to analyze the prediction accuracy of the compressive strength and bonding rate of pellets.The improved algorithm model can predict the compressive strength and bonding rate of pellets.At the same time,the improved algorithm is compared with the naive Bayes algorithm and linear regression algorithm.The results show that the three error values of the improved algorithm are lower than the two comparison algorithms.The model has the shortest training time,good generalization performance,and more accurate and practical predictions for industrial fields such as steel.Figure 29;Table 12;Reference 56...