Interpretable modeling of metallurgical responses for an industrial coal column flotation circuit by XGBoost and SHAP-A “conscious-lab” development

Surprisingly, no investigation has been explored relationships between operating variables and metallurgical responses of coal column flotation (CF) circuits based on industrial databases for under operation plants. As a novel approach, this study implemented a conscious-lab “CL” for filling this gap. In this approach, for developing the CL dedicated to an industrial CF circuit, SHapley Additive exPlanations (SHAP) and extreme gradient boosting (XGBoost) were powerful unique machine learning systems for the first time considered. These explainable artificial intelligence models could effectively convert the dataset to a basis that improves human capabilities for better understanding, reasoning, and planning the unit. SHAP could provide precise multivariable correlation assessments between the CF dataset by using the Tabas Parvadeh coal plant (Kerman, Iran), and showed the importance of solid percentage and washing water on the metallurgical responses of the coal CF circuit. XGBoost could predict metallurgical responses (R-square > 0.88) based on operating variables that showed quite higher accuracy than typical modeling methods (Random Forest and support vector regression).     

Machine learning of phases and mechanical properties in complex concentrated alloys

The mechanical properties of complex concentrated alloys (CCAs) depend on their formed phases and corresponding microstructures.The data-driven prediction of the phase formation and associated mechanical properties is essential to discovering novel CCAs.The present work collects 557 samples of various chemical compositions,comprising 61 amorphous,167 single-phase crystalline,and 329 multi-phases crystalline CCAs.Three classification models are developed with high accuracies to category and understand the formed phases of CCAs.Also,two regression models are constructed to predict the hard-ness and ultimate tensile strength of CCAs,and the correlation coefficient of the random forest regression model is greater than 0.9 for both of two targeted properties.Furthermore,the Shapley additive expla-nation (SHAP) values are calculated,and accordingly four most important features are identified.A significant finding in the SHAP values is that there exists a critical value in each of the top four fea-tures,which provides an easy and fast assessment in the design of improved mechanical properties of CCAs.The present work demonstrates the great potential of machine learning in the design of advanced CCAs.     

基于链上数据的区块链欺诈账户检测研究

针对区块链上存在的欺诈账户给交易带来的安全问题,提出了基于机器学习的欺诈账户的检测及特征分析模型,将以太坊上真实的链上数据进行特征提取后作为模型的数据来源,通过对不同的机器学习方法进行比较得到最优模型并进行迭代训练以获得最佳的预测模型,同时引入 SHAP值对数据特征进行分析。实验结果表明,基于XGBoost的欺诈账户检测模型在RMSE、MAE和R2三组指标上达到了0.205、0.084和0.833,优于其余的对比模型,并结合SHAP值识别出预测欺诈账户的关键因素,为区块链的交易安全提供决策参考。     

基于机器学习的心脏病预测模型研究

心脏病作为当今社会对人类生活健康威胁最严重的血管疾病之一，不仅严重威胁人类的生命安全，而且高昂的治疗费用还会给家庭和社会带来严重的经济负担。针对目前心脏病预测研究中存在准确性不足及缺乏特征可解释性的问题，通过挖掘影响心脏病的重要特征，实现对心脏病准确预测和影响因素可解释性分析。首先，利用T检验来分析特征之间的显著差异性（P-value），通过P-value值降序排列选出特征进行组合。然后，利用十种机器学习模型和SHAP方法实现对心脏病的预测及其特征可解释性分析。在UCI心脏病数据集上进行验证实验，在七个医学领域广泛使用的评价指标上都到达了1，优于与对比论文实验结果。最后，利用SHAP方法对13个特征进行可解释性分析，通过特征重要性排序可视化结果，挖掘单个特征与心脏病之间的关联，能为医生对心脏病的精准医疗提供决策支持。     

基于XGBoost算法的山东烟叶质量预测模型初探

为挖掘烟叶化学成分与感官质量之间的关系,探究机器学习算法在烟叶质量评价领域的应用效果,以山东烟叶为试验材料,开展了常规成分、生物碱、有机酸、多酚和单双糖等20项主要化学成分检测和感官质量评价,并根据感官质量优劣将其划分为好、中、差3个质量档次。利用遗传算法对XGBoost进行超参数寻优,建立了基于化学成分的山东烟叶质量档次预测模型,同时引入SHAP value模型解释框架进行全局解释与特征依赖分析。所建预测模型对山东烟叶质量档次判别准确率为85%,尤其对第3质量档次识别效果最佳。SHAP value全局解释表明,影响山东烤烟质量的7个特征指标贡献度排名为：酸酚比>蔗糖>氯>烟碱>降烟碱>柠檬酸>糖碱比,其中糖碱比、蔗糖、酸酚比分别为好、中、差质量档次判别贡献最大的化学指标。基于XGBoost算法的山东烟叶质量预测模型在烟叶质量档次判别应用中有效、可靠、可解释性强,对于烟叶质量评价和烟叶生产具有一定指导意义。     

一种法律判决预测的影响因素分析方法

法律判决预测是指机器依据案件事实描述,对法律案件的判决罪名进行预测的方法,是人工智能技术在法律领域一项很有前途的应用。随着人工智能领域的迅猛发展,相关技术在法律领域的应用愈加广泛,许多经典的模型在法律判决预测方面取得了不错的结果。机器学习方法效果虽好,但一直无法解决预测结果的解释说理问题,预测结果存在黑盒性特征,无法获得结论的支撑依据。针对上述问题,本文提出一种法律判决预测的影响因素分析方法,结合中文分词技术、支持向量机技术（Support Vector Machine, SVM）和解释预测统一框架（SHapley Additive exPlanations, SHAP）,实现对法律案件判决结果的智能预测,并对预测结果的影响因素进行科学的分析,给出对预测结果影响较大的关键因素,为预测结果提供支撑依据。     

Investigation of feature contribution to shield tunneling-induced settlement using Shapley additive explanations method

Accurate prediction of shield tunneling-induced settlement is a complex problem that requires consideration of many influential parameters. Recent studies reveal that machine learning (ML) algorithms can predict the settlement caused by tunneling. However, well-performing ML models are usually less interpretable. Irrelevant input features decrease the performance and interpretability of an ML model. Nonetheless, feature selection, a critical step in the ML pipeline, is usually ignored in most studies that focused on predicting tunneling-induced settlement. This study applies four techniques, i.e. Pearson correlation method, sequential forward selection (SFS), sequential backward selection (SBS) and Boruta algorithm, to investigate the effect of feature selection on the model's performance when predicting the tunneling-induced maximum surface settlement (S_max). The data set used in this study was compiled from two metro tunnel projects excavated in Hangzhou, China using earth pressure balance (EPB) shields and consists of 14 input features and a single output (i.e. S_max). The ML model that is trained on features selected from the Boruta algorithm demonstrates the best performance in both the training and testing phases. The relevant features chosen from the Boruta algorithm further indicate that tunneling-induced settlement is affected by parameters related to tunnel geometry, geological conditions and shield operation. The recently proposed Shapley additive explanations (SHAP) method explores how the input features contribute to the output of a complex ML model. It is observed that the larger settlements are induced during shield tunneling in silty clay. Moreover, the SHAP analysis reveals that the low magnitudes of face pressure at the top of the shield increase the model's output.     

一种基于神经网络的 THD 参数快速估计算法

模数转换器(ADC)测试主要包括静态参数和动态参数两个测试过程。随着性能的提升,ADC的测试复杂度和成本也急剧增加。替代测试,即通过分析两类参数间的关系来实现一个测试过程得到两类参数,已被证明是降低ADC测试复杂度和成本的主要方案之一。本文通过构建基于人工神经网络的参数预测模型来实现替代测试,模型以总谐波失真为预测目标,以静态性能参数为输入特征。针对高维的ADC非线性曲线,文章结合统计分析和主成分分析设计了专用的特征提取方法,在降低特征维度的同时尽可能地减少了信息损失。模型在测试集上的预测结果与参考值的均方误差和拟合优度分别达到了1.15 dB和0.6,显著优于相关对比模型。此外,在SHAP解释器的框架下分析了上述模型的预测目标和特征变量之间的依赖关系,并得到了有意义的结果。     

融合XGBoost与SHAP的冠心病预测及其特征分析模型

为了解决冠心病诊断模型中性能无法满足临床应用要求、缺乏可解释性的问题,提出一种融合XGBoost与SHAP的冠心病预测及其特征分析模型。在对数据集进行特征工程的基础上,将处理好的数据集输入XGBoost模型进行训练,并且对模型进行优化,进一步提高了模型的性能表现;其次,与基于SVM、朴素贝叶斯等六种机器学习模型以及八种主流机器学习模型进行实验对比,参数优化后的XGBoost模型在准确率、特异度、F1值和AUC值四个指标上分别达到0.9942、0.9970、0.9941和0.9998,均优于已有模型;最后引入SHAP框架增强模型可解释性,综合四种模型特征重要性排序结果,识别出影响冠心病的重要因素,为医生作出正确的诊断提供决策参考。     

土石坝渗流性态分析的IAO-XGBoost集成学习模型与 预测结果解释

针对现有土石坝渗流数值模拟方法计算效率较低、难以实时分析大坝渗流性态，而现有基于机器学习算法建立的代理模型又存在模型可解释性较差的问题，提出土石坝渗流性态分析的IAO-XGBoost集成学习模型，并基于Shapley加性解释(SHapley Additive exPlanation, SHAP)理论对预测结果进行解释。在采用多地质体自动建模方法和CFD技术对大坝渗流场进行计算分析的基础上，基于改进的天鹰(Improved Aquila Optimization, IAO)算法优化极限梯度提升(eXtreme Gradient Boosting, XGBoost)集成学习算法中的n＿estimators、max＿depth和learning＿rate等超参数，进而建立基于IAO-XGBoost集成学习算法的大坝渗流性态指标预测模型，以揭示上下游水位和坝基地层渗透系数等输入特征变量与渗流性态指标模拟值间的复杂非线性映射关系。进一步地，将IAO-XGBoost集成学习算法与可解释机器学习框架SHAP理论相结合，挖掘影响大坝渗流性态指标预测结果的关键特征，并解释特征变量对渗流性态指标预测的影响...