基于集成LSSVM增量学习方法研究与应用

Research and application of incremental learning method based on integrated LSSVM

针对最小二乘支持向量机(LSSVM)模型在测量精度及泛化能力上的不足,及不能准确预测新工况数据的特点,提出基于集成LSSVM增量学习的模型。通过单个LSSVM模型的建立,构建集成LSSVM增量学习模型,在增量学习方法基础上加入样本判别规则,使模型能及时更新新工况数据。运用集成LSSVM增量学习模型与集成LSSVM模型对溶剂油分离过程的120号溶剂油流量数据进行预测分析,所提模型预测效果显著提高,验证了其有效性。

The LSSVM model has disadvantages of the poor measurement accuracy and generalization ability,and it can not accurately predict the characteristics of the new working data.A model of incremental learning based on integrated LSSVM was proposed.Through the establishment of a single LSSVM model,an integrated LSSVM incremental learning model was built.Based on incremental learning method,the sample discriminant rules were introduced,so that the model updated the new data.The integrated LSSVM incremental learning model and the integrated LSSVM model were integrated to predict and analyze the solvent oil flow data of 120 solvent oil separation process.The effectiveness of the proposed method is improved significantly,and the effectiveness of the proposed method is verified.