关键词： NIR光谱   咖啡   蛋白质   SVM   ELM   潜变量技术  

摘要： 采用近红外(NIR)光谱快检技术实现对咖啡蛋白质的定量检测,研究支持向量机(SVM)和极限学习机(ELM)等机器学习方法在建模分析中的实用性。结合潜变量分析技术,建立潜变量SVM(LV-SVM)模型和潜变量ELM(LV-ELM)模型,通过调试潜变量个数和机器学习关键参数的联合优选,实现数据降维和机器学习关键参数的同过程优化。运用定标—验证—测试机制,利用定标集样本建立咖啡蛋白质的NIR分析模型,随参数变动形成三维随动优选结构的建模预测结果,结合验证集样本对模型进行联合优选,然后将优化模型应用于测试集样本进行模型评价。LV-SVM建模优选的验证集预测均方根误差为6.797,对应的测试集预测均方根误差为8.384。LV-ELM建模优选的验证集预测均方根误差为6.118,对应的测试集预测均方根误差为7.837。与常规偏最小二乘(PLS)方法相比较,LV-SVM和LV-ELM方法均取得更好的预测结果,验证了潜变量机器学习方法在近红外定量分析中的应用优势,该方法有望应用于不同类型的咖啡各成分含量检测。

关键词： micro-milling   tool condition monitoring   machine learning   random forest   machining processes   micro- and nano-machining and processing   precision and ultra-precision machining   sensing   monitoring and diagnostics  

摘要： Tool condition monitoring is difficult in micro-milling due to irregular wear and chipping of the cutting edges, which lead to unexpected tool breakage. This study demonstrates the use of force data to reliably predict different tool life stages until tool breakage, while micro-milling hard materials like stainless steel (SS304) using tungsten carbide tools of 500 mu m diameter. Extensive experiments involving machining of 465 slots over 62 min of machining time were performed in this study. The resulting voluminous force data were analyzed to divide the tool life into three stages based on the variation in the forces and other related features. The first stage is the initial 12.5% of the tool life, second stage consists of 12.5-70% of tool life, and the third stage is from 70% to 100% tool life. The analysis of the tool wear and cutting forces shows that the average tool diameter reduces by 32 mu m, 67 mu m and 108 mu m, and the average resultant cutting force were 2.45 N, 4.17 N, and 4.93 N in stage 1, 2, and 3, respectively. To avoid catastrophic breakage of the tool, the tool life stages are predicted from the force data using machine learning models. Among the machine learning models, random forest method gave a better prediction accuracy of 88.5%. The model was further improved by incorporating the initial cutting edge radius as an additional feature, and the variance in the prediction was seen to drop by 48.76%.

关键词： Machine learning   Dimensionality   Small sample size   High correlation   Peripheral neuropathy   Carpal tunnel syndrome  

摘要： Peripheral neuropathy that interrupts sensorimotor integration in the motor control process will lead to hand and upper limb motor function deficits in daily life. However, behavioral biomechanics and motor functions have never been considered in available diagnoses and clinical evaluations. Previous studies that investigate the behavioral biomechanics to delineate a specific peripheral neuropathy and its severity have shown evidences that certain biomechanical parameters have the potential to be identified as biomarkers for the detection of the neuropathy from an early stage. Nevertheless, datasets formed by behavioral biomechanical parameters are often characterized by the high dimensionality, the small sample size, and the high redundancy, which brings us challenges for making binary classification between patients and healthy controls. We propose a two-phase machine learning protocol using Random Forests (RFs) for the early variable screening and the (K)PCA-SVM system for the prediction and the final identification of biomarkers. We apply the proposed protocol to an example application of Carpal Tunnel Syndrome (CTS) and its prediction accuracy reaches 90.3% with 6 biomarker variables identified from 700 initial input variables. These promising results provide a paradigm shift of guidelines and directions of clinical test designs toward novel diagnostic optimization in the future.

关键词： 丙型肝炎   丙肝病毒感染   医疗水平   确诊率   献血者   损伤程度  

摘要： 以目前的医疗水平丙肝病毒可以确诊并根治,但是丙肝病毒并没有被彻底消灭,主要是因为丙肝病毒感染病例的确诊率很低,有许多被感染而未确诊的患者,因此,对丙型肝炎进行早期的检测则显得尤为重要。血清丙氨酸氨基转移酸(ALT)和天门冬氨酸氨基转移酶(AST)水平在一定程度上反映了肝脏的损伤程度,检测他们,可以对献血者进行初步的丙型肝炎筛查。

关键词： 创伤静脉   血栓栓塞症   机器学习   随机森林  

摘要： 目的:筛选住院病历系统中可能与静脉血栓栓塞症相关的特征,与Caprini评分相结合,采用机器学习方法进行建模,在量表评估的基础上进一步提升VTE预测性能。方法:回顾了2019年9月至2020年3月在四川大学华西医院创伤医学中心住院的903例创伤患者,基于Lasso回归方法进行特征筛选后,结合随机森林算法构建VTE预测模型,并与单独使用Caprini评分的预测效果进行比较。结果:Caprini评分在预测静脉血栓栓塞事件上表现出了较好的区分度(TPR=0.667,FPR=0.227,AUC=0.773),最终构建的机器学习模型在Caprini评分的基础上,增加4个预测特征,可进一步提升预测效果(TPR=0.757,FPR=0.290,AUC=0.799)。结论:结合Caprini评分及生理生化特征的机器学习模型适用于对住院创伤患者的静脉血栓栓塞症风险预测。

摘要： Genomic-wide selection (GWS) consists of the use of a large number of molecular markers for the prediction of genetic values and has been shown to be highly relevant for genetic improvement. The objective of this work was to evaluate and compare the predictive performance of statistical (ridge regression-best linear unbiased predictor [RR-BLUP] and BayesB) and machine learning methods through GWS in simulated populations with traits presenting different levels of heritability and quantitative trait loci (QTL) numbers in the presence of dominant and epistatic effects. The simulated genome of population F-2 was formed by 1,000 individuals and genotyped with 2,010 single nucleotide polymorphism (SNP) markers. Twenty-six traits were simulated considering QTL numbers ranging from two to 88 and heritabilities of .3 and .6. The selective and predictive performances were evaluated using the multilayer perceptron (MLP), radial basis function (RBF), decision trees (DT), bagging (BA), random forest (RF), and boosting (BO) machine learning models and the classical RR-BLUP and BayesB methods. A high effect of heritability was observed for the results of selective accuracy when compared to the increased QTL number. In addition, the selective accuracy based on the number of QTL demonstrates that the application of alternative machine learning models, such as RBF, BA, BO, and RF, can be suitable for the analysis according to QTL number. Machine learning methods are powerful tools for predicting genetic values with epistatic gene control in traits with different degrees of heritability and different numbers of controlling genes.

关键词： Hesitancy   Airport choice   Multi-airport region   Multinomial logit   Random forest   Deep reinforcement learning  

摘要： Noticing that people may become more hesitant about their airport choice when exposed to abundant information and multi-channel on-line sales opportunities, this paper analyzes hesitancy in the choice of transfer airport in a multi-airport region using three approaches: a multinomial logit model, a random forests algorithm, and deep reinforcement learning. These methods are applied to the outcomes of a stated choice experiment, which was designed to survey the iterative airport choice process under hesitancy. The results indicate that a quarter of the passengers show evidence of hesitancy. Further, the deep reinforcement learning approach is found to provide the highest accuracy in representing the stated choices;the random forests algorithm also gives good results in capturing the reconsideration decision, while the multinomial logit model seems less accurate in representing the choice behavior of interest. Moreover, hesitancy involved a switch from flight-related attributes to ground transport-related attributes. Managerial implications for airlines and airports are discussed.

关键词： biological manufacturing process   low pH viral inactivation   machine learning   monoclonal antibody   pathogen safety  

摘要： The production of recombinant therapeutic proteins from animal or human cell lines entails the risk of endogenous viral contamination from cell substrates and adventitious agents from raw materials and environment. One of the approaches to control such potential viral contamination is to ensure the manufacturing process can adequately clear the potential viral contaminants. Viral clearance for production of human monoclonal antibodies is achieved by dedicated unit operations, such as low pH inactivation, viral filtration, and chromatographic separation. The process development of each viral clearance step for a new antibody production requires significant effort and resources invested in wet laboratory experiments for process characterization studies. Machine learning methods have the potential to help streamline the development and optimization of viral clearance unit operations for new therapeutic antibodies. The current work focuses on evaluating the usefulness of machine learning methods for process understanding and predictive modeling for viral clearance via a case study on low pH viral inactivation.

关键词： Decision model   Machine learning   Neovascular age-related macular degeneration   Non-interventional study   Predictive analytics   Real world evidence  

摘要： Purpose: The effectiveness of intravitreal anti-vascular endothelial growth factor agents is usually lower in real world settings compared with randomized clinical trials (RCTs), often limiting the use of real-world evidence (RWE) in regulatory and healthcare decisions. The current analysis aimed to develop and validate an algorithm to explain the difference in outcomes between RWE studies and RCTs in patients with neovascular age-related macular degeneration. Methods: The algorithm was developed using ranibizumab real world data (RWD) from the US and validated on Australian and UK RWD. A decision model was developed using machine learning principles, in which the model learns how to partition the most influential factors (out of 59 variables) so that they maximally relate to the change in visual acuity (VA) over 12 months. Results: The algorithm identified baseline VA <73 Early Treatment Diabetic Retinopathy Study letters, presence of baseline subretinal fluid, and administration of three loading doses by Day 90 from drug initiation as the characteristics with the greatest impact on VA at month 12. When applying the different criteria, RWE outcomes became similar to those obtained in known RCTs. Conclusion: Machine learning techniques can be used to classify real world cohorts and identify subsets of patients who benefit to the same extent as that reported in RCTs. This methodology may support the translation of clinical trial findings to treatment performance in the clinical practice setting.

关键词： postural control   deep learning   machine learning   joint–   node plot  

摘要： Postural control decreases with aging. Thus, an efficient and accurate method of detecting postural control is needed. We enrolled 35 elderly adults (aged 82.06 +/- 8.74 years) and 20 healthy young adults (aged 21.60 +/- 0.60 years) who performed standing tasks for 40 s, performed six times. The coordinates of 15 joint nodes were captured using a Kinect device (30 Hz). We plotted joint positions into a single 2D figure (named a joint-node plot, JNP) once per second for up to 40 s. A total of 15 methods combining deep and machine learning for postural control classification were investigated. The accuracy, sensitivity, specificity, positive predicted value (PPV), negative predicted value (NPV), and kappa values of the selected methods were assessed. The highest PPV, NPV, accuracy, sensitivity, specificity, and kappa values were higher than 0.9 in validation testing. The presented method using JNPs demonstrated strong performance in detecting the postural control ability of young and elderly adults.