一种基于密度不平衡样本数据的材料性能预测方法及系统

• Format: Patent
• Language: Chinese
• Published: 26.04.2024
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTEDFOR SPECIFIC APPLICATION FIELDS; PHYSICS

本发明涉及一种基于密度不平衡样本数据的材料性能预测方法及系统。本发明首先定位第一材料数据集和第二材料数据集的边界样本，然后利用原始的样本数据集训练第一材料分类模型，利用边界样本训练第二材料分类模型，进而将第一材料分类模型和第二材料分类模型进行融合，利用融合后的集成模型进行材料性能的预测分类，实现了边界样本的定位，并通过基于边界样本的单独训练，提升了对少数类样本预测的准确性。 The invention relates to a material performance prediction method and system based on density imbalance sample data. The method comprises the following steps: firstly, positioning boundary samples of a first material data set and a second material data set, then training a first material classification model by using an original sample data set, training a second material classification model by using the boundary samples, and fusing the first material classification model and the second material classification model; and using the integrated model after fusion to predict and classify the material performance. Positioning of boundary samples is achieved, and through independent training based on the boundary samples, the accuracy of predicting minority class samples is improved.