本发明公开了一种高性能新型合金的预测方法及系统,方法包括:获取历史合金数据集;对历史合金数据集进行数据预处理,得到每一种历史合金的合金元素特征因子集;通过相关性分析法、梯度下降法及SHAP法对合金元素特征因子集进行筛选,得到关键特征因子组合;利用奥图纳算法根据数据情况构建出大批量不同的模型超参数组合,结合关键特征因子组合进行神经网络模型超参数自优化和模型自筛选,根据模型泛化能力和测试准确率来判断是否得到最佳模型;获取元素周期表中所有元素的元素特征因子数据;根据元素特征因子数据及最佳模型预测得到高性能新型合金。解决了手动调整超参数过程机器学习模型泛化能力弱,模型过拟合或欠拟合导致的准确率低的问题。
The present invention discloses a prediction method and system for a high-performance novel alloy, comprising: obtaining a historical alloy dataset; Preprocess the historical alloy dataset to obtain the characteristic factor set of alloy elements for each historical alloy; By using correlation analysis, gradient descent method, and SHAP method to screen the feature factor set of alloy elements, key feature factor combinations are obtained; Using the Otuna algorithm to construct a large number of different model hyperparameter combinations based on data conditions, combined with key feature factor combinations for neural network model hyperparameter self optimization and model self screening, to determine whether the optimal model is obtained based on the model's generalization ability and testing accuracy; Obtain element characteristic factor data for all elements in the periodic table; Based on element characteristic factor data and the best model prediction, high-performance new alloys are obtained. Solved the problem of weak generalization ability of machine learning models and low accuracy caused by overfitting or underfitting during manual adjustment of hyperparameters.