用于早期火灾分类的非线性决策树支持向量机

对早期火灾信息进行研究,提出了一种基于非线性决策树的支持向量机多类分类模型。该模型利用非线性映射将样本投影到高维特征空间,比较每类样本在高维空间的分布情况,进行聚类构造出一个二叉决策树,使容易区分的类别从根节点逐层分类出来,有效克服了错分积累和避免不可分情况;同时,各个节点采用二值最小二乘小波支持向量机,以获得较高的泛化能力。该文将该模型用于早期火灾分类,并与BP神经网络、K近邻法和决策树方法进行比较,实验结果表明,该模型对早期火灾的识别率更高。     

Monitoring in adaptive co-management: Toward a learning based approach

The recognition of complexity and uncertainty in natural resource management has lead to the development of a wealth of conceptual frameworks aimed at integrated assessment and complex systems monitoring. Relatively less attention has however been given to methodological approaches that might facilitate learning as part of the monitoring process. This paper reviews the monitoring literature relevant to adaptive co-management, with a focus on the synergies between existing monitoring frameworks, collaborative monitoring approaches and social learning. The paper discusses the role of monitoring in environmental management in general, and the challenges posed by scale and complexity when monitoring in adaptive co-management. Existing conceptual frameworks for monitoring relevant to adaptive co-management are reviewed, as are lessons from experiences with collaborative monitoring. The paper concludes by offering a methodological approach to monitoring that actively seeks to engender reflexive learning as a means to deal with uncertainty in natural resource management.     

The effect of workers' visibility on effectiveness of intervention programs: supervisory-based safety interventions

INTRODUCTION: This paper discusses an organizational change intervention program targeting safety behaviors and addresses important considerations concerning the planning of organizational change. Using layout of the plant as a proxy for ease of daily leader-member interaction, the effect of workers' visibility on the effectiveness of supervisory-based safety (SBS) interventions is examined. Through a reinforcement-learning framework, it is suggested that visibility can affect supervisors' incentive to interact with subordinates regarding safety-related issues. METHOD: Data were collected during SBS intervention studies in five manufacturing companies. RESULTS: Results suggest a reinforcement cycle model whereby increased visibility generates more frequent exchanges between supervisors and employees, resulting in improved safety behavior among employees. In turn, employees' safer behavior reinforces continued supervisory safety-related interaction. CONCLUSION AND IMPACT ON INDUSTRY: Visibility is an important moderator in supervisory based safety interventions, and can serve to increase workplace safety. Implications of these findings for safety are discussed.     

一种新型的矿井突水分析与预测的支持向量机模型

针对矿井突水样本数少,信息不完整的特点,提出了矿井突水分析的线性核H-SVMs模型。推导模型的理论推广误差公式,设计自顶向下基于SVM最大间隔逐层分类构造H-SVMs的新方法,并应用于实际的矿井突水预测。实验结果表明,线性核H-SVMs模型结构简单、泛化能力强,不仅能很好地预测矿井突水,而且其层次结构能正确反映突水的等级关系,各判别函数的法向量还可以指示各突水影响因素的权重,通过判决函数能有效分析突水影响因素并提取突水预测规则,为矿井突水预测提供了新的方法。     

基于车载机器视觉的汽车安全技术

在描述汽车驾驶操作过程基础上,根据信息获取范围将汽车安全辅助驾驶的机器视觉分为外部信息的机器视觉与内部信息的机器视觉技术,包括视觉增强、视野扩展、道路环境理解、视线跟踪与驾驶疲劳监测;逐个阐述其作用机理,综述其国内外研究现状,并归纳出研究重点;分析了汽车安全辅助驾驶系统中机器视觉技术当前研究不足。笔者认为,低能见度驾驶员视觉增强方法、道路环境理解信息融合以及驾驶疲劳检测等技术需进一步开展研究。     

河北省高校社会体育专业培养目标及学生满意度调查分析

运用文献资料法、问卷调查法、数理统计法和对比分析法,对我省社会体育专业本专科院校的培养目标情况、学生收获和满意度情况进行调查分析。研究结果表明,中环院社会体育专业学生对课程体系设置的满意度最高,就业信心指数最大。进一步分析可知,以社会体育产业为目标定位、以能力本位为教育模式对社会体育专业学生的培养具有一定的优越性     

高校传统诚信教育的理性重建

"诚信"是中华民族的道德规范,我国大学教育一直秉承诚信教育传统,为培养诚实守信的公民而不懈努力。但面对市场经济带来的思想、观念、制度冲击,传统诚信教育的弊端日益凸现,如何看待诚信、如何促进诚信教育的良性发展、如何保障诚信教育的切实践行等问题,应该成为高等教育者们不断思考的焦点,而重建传统诚信教育要通过建设校园诚信文化和构建校长负责制等措施来实现。     

Social learning from public engagement: dreaming the impossible?

Learning that transcends participation processes is critical if public engagement is to translate into a legacy of enhanced environmental citizenship. However, a lack of empirical evidence has limited discussion to date to largely ‘aspirational’ claims. This paper offers the first rigorous examination of whether public participation does generate beyond-process social learning. Initially we review the literature on public participation and environmental citizenship to identify the key dimensions of social learning. We then re-visit a well-worked case study of an innovative public engagement process on the Hampshire waste strategy from the 1990s. Approximately one third of the original participants have been interviewed to identify whether and how the experience had a lasting effect on them. Key methodological difficulties are discussed, not least the analytical difficulties of attributing learning to a process that happened ten years previously. However, we argue that there is evidence that both instrumental and communicative learning have taken place, and conclude by identifying key areas that require further research.     

基于支持向量机的浮游植物密度预测研究

利用支持向量机方法对具有非线性及突发性特点的浮游植物密度进行了预测,同时与人工神经网络方法预测的结果进行了比较.结果表明,无论是拟和能力还是预测能力,支持向量机方法都明显优于人工神经网络方法,支持向量机方法比较适合于具有小样本、非线性特点的浮游植物密度预测研究.     

基于PCA-PNN的采空区多源指标危险性辨识

为了提高采空区多源指标危险性辨识的预测精度,基于主成分分析(PCA)和概率神经网络(PNN),提出1种采空区多源指标危险性辨识方法。将影响华东某地区矿山采空区危险性辨识的9项因素作为主要影响因素,并以96个实测采空区为例进行分级。研究结果表明：与朴素贝叶斯、随机森林和AdaBoost 3种机器学习算法相比,PNN在测试集上表现更好,对实际工程具有良好的指导意义和应用价值。