Validity of the Acti4 method for detection of physical activity types in free-living settings: comparison with video analysis

This study examined the ability of the Acti4 software for identifying physical activity types from accelerometers during free-living with different levels of movement complexity compared with video observations. Nineteen aircraft cabin cleaners with ActiGraph GT3X+ accelerometer at the thigh and hip performed one semi-standardised and two non-standardised sessions (outside and inside aircraft) with different levels of movement complexity during working hours. The sensitivity for identifying different activity types was 75.4–99.4% for the semi-standardised session, 54.6–98.5% outside the aircraft and 49.9–90.2% inside the aircraft. The specificity was above 90% for all activities, except ‘moving’ inside the aircraft. These findings indicate that Acti4 provides good estimates of time spent in different activity types during semi-standardised conditions, and for sitting, standing and walking during non-standardised conditions with normal level of movement complexity. The Acti4 software may be a useful tool for researchers and practitioners in the field of ergonomics, occupational and public health.

Practitioner Summary: Being inexpensive, small, water-resistant and without wires, the ActiGraph GT3X+ by applying the Acti4 software may be a useful tool for long-term field measurements of physical activity types for researchers and practitioners in the field of ergonomics, occupational and public health.     

服务功能链中基于机器学习的QoE评估与预测

在软件定义网络与网络功能虚拟化协同的网络架构下,只考虑单个服务质量（QoS）指标的服务功能链部署无法满足用户的多业务体验需求。提出一种基于机器学习的服务功能链部署模型。基于层次分析法构造MPNQ2算法以建立QoS与体验质量（QoE）的映射关系,得出影响QoE的网络参数并评估其影响权重。在此基础上,利用具备较强综合学习和泛化能力的随机森林模型对服务功能链的QoE进行预测。实验结果表明,与梯度提升决策树、线性判别分析等机器学习模型相比,随机森林模型为预测QoE的最佳模型,同时在影响QoE的网络参数中,丢包率对服务功能链的部署影响最大。     

Combining bagging,boosting, rotation forest and random subspace methods

Bagging, boosting, rotation forest and random subspace methods are well known re-sampling ensemble methods that generate and combine a diversity of learners using the same learning algorithm for the base-classifiers. Boosting and rotation forest algorithms are considered stronger than bagging and random subspace methods on noise-free data. However, there are strong empirical indications that bagging and random subspace methods are much more robust than boosting and rotation forest in noisy settings. For this reason, in this work we built an ensemble of bagging, boosting, rotation forest and random subspace methods ensembles with 6 sub-classifiers in each one and then a voting methodology is used for the final prediction. We performed a comparison with simple bagging, boosting, rotation forest and random subspace methods ensembles with 25 sub-classifiers, as well as other well known combining methods, on standard benchmark datasets and the proposed technique had better accuracy in most cases.     

Examining runner's outdoor heat exposure using urban microclimate modeling and GPS trajectory mining

It is important to quantify human heat exposure in order to evaluate and mitigate the negative impacts of heat on human well-being in the context of global warming. This study proposed a human-centric framework to examine human personal heat exposure based on anonymous GPS trajectories data mining and urban microclimate modeling. The mean radiant temperature (T_mrt) that represents human body's energy balance was used to indicate human heat exposure. The meteorological data and high-resolution 3D urban model generated from multispectral remotely sensed images and LiDAR data were used as inputs in urban microclimate modeling to map the spatio-temporal distribution of the T_mrt in the Boston metropolitan area. The anonymous human GPS trajectory data collected from fitness Apps was used to map the spatio-temporal distribution of human outdoor activities. By overlaying the anonymous GPS trajectories on the generated spatio-temporal maps of T_mrt, this study further examined the heat exposure of runners in different age-gender groups in the Boston area. Results show that there is no significant difference in terms of heat exposure for female and male runners. The female runners in the age of 45–54 are exposed to more heat than female runners of 18–24 and 25–34, while there is no significant difference among male runners. This study proposed a novel method to estimate human heat exposure, which would shed new light on mitigating the negative impacts of heat on human health.     

Preprocessing techniques for context recognition from accelerometer data

The ubiquity of communication devices such as smartphones has led to the emergence of context-aware services that are able to respond to specific user activities or contexts. These services allow communication providers to develop new, added-value services for a wide range of applications such as social networking, elderly care and near-emergency early warning systems. At the core of these services is the ability to detect specific physical settings or the context a user is in, using either internal or external sensors. For example, using built-in accelerometers, it is possible to determine whether a user is walking or running at a specific time of day. By correlating this knowledge with GPS data, it is possible to provide specific information services to users with similar daily routines. This article presents a survey of the techniques for extracting this activity information from raw accelerometer data. The techniques that can be implemented in mobile devices range from classical signal processing techniques such as FFT to contemporary string-based methods. We present experimental results to compare and evaluate the accuracy of the various techniques using real data sets collected from daily activities.     

Discriminating the early stages of Sirex noctilio infestation using classification tree ensembles and shortwave infrared bands

The woodwasp Sirex noctilio is causing extensive damage to Pinus patula trees in the summer rainfall areas of South Africa. The ability to remotely detect S. noctilio infestation remains crucial for monitoring purposes and for the effective deployment of suppression activities. In this study, we evaluated whether random forest and boosting trees can accurately discriminate between healthy trees and the early stages of S. noctilio infestation using reflectance measurements in the shortwave infrared (SWIR). Three variable selection methods, namely, a filter, the random forest out-of-bag samples and a wrapper algorithm, were used to select the smallest subset of SWIR bands. The results show that random forest produces better classification results than the competing boosting trees for all three variable selection methods, even when noise is introduced into the SWIR bands and class labels. The ability of the bands centred at 1990, 2009, 2028, 2047 and 2065 nm to discriminate between healthy trees and the early stages of infestation could be explained due to the rapid physiological changes that occur as a result of the toxic mucus and a fungus that S. noctilio injects into the tree. Overall, the results are encouraging and show that there is a link between the selected SWIR bands and existing physiological knowledge, thereby improving the chances of detecting the early stages of S. noctilio infestation at a canopy or landscape level.     

Mobile health monitoring system based on activity recognition using accelerometer

We propose a new method to recognize a user’s activities of daily living with accelerometers and RFID sensor. Two wireless accelerometers are used for classification of five human body states using decision tree, and detection of RFID-tagged objects with hand movements provides additional instrumental activity information. Besides, we apply our activity recognition module to the health monitoring system. We derive linear regressions for each activity by finding the correlations between the attached accelerometers and the expended calories calculated from gas exchange analyzer under different activities. Finally, we can predict the expended calories more efficiently with only accelerometer sensor depend on the recognized activity. We implement our proposed health monitoring module on smart phones for better practical use.     

The effect of physical training on objective and subjective measures of productivity and efficiency in industry

The aims of this study were to investigate the associations between an activity logbook and the RT3 accelerometer and to assess whether the RT3 can discriminate activity levels in healthy adults. Ten participants completed two trials wearing an RT3 accelerometer over a 4–6 h period and completed a detailed activity log. Results showed a poor correlation between the RT3 in moderate activities (r = 0.22) in comparison to low (r = 0.52) and hard (r = 0.70) from the logbook. A significant difference was found in average RT3 vector magnitude (VM) counts/min in each activity level (p < 0.0001). Discriminant analysis demonstrated that an RT3VM counts/min value of approximately 500 was found to have high sensitivity (88%), and specificity (88%) for discriminating between low and moderate activity levels from the logbook. This study found that accelerometry has the potential to discriminate activity levels in free living. This study is the first to investigate whether tri-axial accelerometry can discriminate different levels of free-living activity recorded in an activity logbook. The RT3 accelerometer can discriminate between low and moderate physical activities and offers a methodology that may be applicable to future research in occupational settings.     

Linear Programming Boosting via Column Generation

We examine linear program (LP) approaches to boosting and demonstrate their efficient solution using LPBoost, a column generation based simplex method. We formulate the problem as if all possible weak hypotheses had already been generated. The labels produced by the weak hypotheses become the new feature space of the problem. The boosting task becomes to construct a learning function in the label space that minimizes misclassification error and maximizes the soft margin. We prove that for classification, minimizing the 1-norm soft margin error function directly optimizes a generalization error bound. The equivalent linear program can be efficiently solved using column generation techniques developed for large-scale optimization problems. The resulting LPBoost algorithm can be used to solve any LP boosting formulation by iteratively optimizing the dual misclassification costs in a restricted LP and dynamically generating weak hypotheses to make new LP columns. We provide algorithms for soft margin classification, confidence-rated, and regression boosting problems. Unlike gradient boosting algorithms, which may converge in the limit only, LPBoost converges in a finite number of iterations to a global solution satisfying mathematically well-defined optimality conditions. The optimal solutions of LPBoost are very sparse in contrast with gradient based methods. Computationally, LPBoost is competitive in quality and computational cost to AdaBoost.     

基于机器学习的TLS恶意加密流量检测方案

首先介绍了安全传输层（TLS,transport layer security）协议的特点、流量识别方法;然后给出了一种基于机器学习的分布式自动化的恶意加密流量检测体系;进而从 TLS 特征、数据元特征、上下文数据特征3个方面分析了恶意加密流量的特征;最后,通过实验对几种常见机器学习算法的性能进行对比,实现了对恶意加密流量的高效检测。     

Heart rate monitoring,activity recognition,and recommendation for e-coaching

Equipped with hardware, such as accelerometer and heart rate sensor, wearables enable measuring physical activities and heart rate. However, the accuracy of these heart rate measurements is still unclear and the coupling with activity recognition is often missing in health apps. This study evaluates heart rate monitoring with four different device types: a specialized sports device with chest strap, a fitness tracker, a smart watch, and a smartphone using photoplethysmography. In a state of rest, similar measurement results are obtained with the four devices. During physical activities, the fitness tracker, smart watch, and smartphone measure sudden variations in heart rate with a delay, due to movements of the wrist. Moreover, this study showed that physical activities, such as squats and dumbbell curl, can be recognized with fitness trackers. By combining heart rate monitoring and activity recognition, personal suggestions for physical activities are generated using a tag-based recommender and rule-based filter.     

Aerobic activity monitoring: towards a long-term approach

With recent progress in wearable sensing, it becomes reasonable for individuals to wear different sensors all day, and thus, global activity monitoring is establishing. The goals in global activity monitoring systems are among others to tell the type of activity that was performed, the duration and the intensity. With the information obtained this way, the individual’s daily routine can be described in detail. One of the strong motivations to achieve these goals comes from healthcare: To be able to tell if individuals were performing enough physical activity to maintain or even promote their health. This work focuses on the monitoring of aerobic activities and targets two main goals: To estimate the intensity of activities, and to identify basic/recommended physical activities and postures. For these purposes, a dataset with 8 subjects and 14 different activities was recorded, including the basic activities and postures, but also examples of household (ironing, vacuum cleaning), sports (playing soccer, rope jumping), and everyday activities (ascending and descending stairs). Data from 3 accelerometers—placed on lower arm, chest, and foot—and a heart rate monitor were analyzed. This paper presents the entire data processing chain, analyses and compares different classification techniques, concerning also their feasibility for portable online activity monitoring applications. Results are presented with different combinations of the sensors. For the intensity estimation task, using the sensor setup composed of the chest accelerometer and the HR-monitor is considered the most efficient, achieving a performance of 94.37 %. The overall performance on the activity recognition task, using all available sensors, is 90.65 % with boosted decision trees—the classifier achieving the best classification results within this work.     

基于Spark平台的网络游戏用户流失预测方法

随着移动互联网的广泛普及,国内网络游戏市场日趋饱和,游戏公司获得新用户的成本不断增加,如何预防存量用户的流失已经成为市场营销的重心。提出了一种基于Spark平台的网络游戏用户流失预测方法,基于一个真实游戏日志数据对用户进行了流失预测。首先,从日志数据中抽取和计算了用户特征;随后,按权重选取了一组重要特征;最后,以特征为输入、流失与否为输出进行了二分类建模。综合比较了随机森林、支持向量机、多层感知机、梯度提升决策树和逻辑回归等6种常见分类算法。实验结果表明,随机森林算法表现最优,模型预测精度达到91%。     

Identifying manufacturing operational conditions by physics-based feature extraction and ensemble clustering

Manufacturing processes usually exhibit mixed operational conditions (OCs) due to changes in process/tool/equipment health status. Undesired OCs are direct causes of out-of-control production and thus need to be identified. Data-driven OC identification has been widely used for recognizing undesired OCs, yet most methods of this kind require labels indicating the OCs in model training. In industrial applications, such labels are rarely available due to delay, incompleteness or physical constraints in data collection. A typical case is the thermal images acquired by in-process infrared camera and pyrometer, which contain rich information about process health status but are unlabeled. To facilitate data-driven OC identification with unlabeled thermal images, this study proposes a feature extraction-clustering framework that characterizes the heat-affected zone by its temperature profile and performs ensemble clustering on the extracted features to label the data. Domain knowledge from plant manufacturing is incorporated in the framework to map cluster labels to OCs. Both offline OC recovery and online OC identification are studied. Thermal images from hot stamping in automotive manufacturing are used to demonstrate and validate the proposed method. The feasibility, effectiveness and generality are well justified by the case study results.     

基于约束条件随机场的Web对象属性标注

条件随机场模型是目前处理We b对象属性标注问题的最佳统计模型。为解决条件随机场模型不能充分利用We b对象和属性标签之间的特征关系这一问题,提出了一种增强约束条件随机场模型。借鉴最大间隔的思想,在原有条件随机场模型中增加约束条件和增强因子以提高模型标注正确率。使用最大似然参数估计方法估计模型特征函数的权重参数,并用Viterbi算法进行预测。在数据集中引入验证集的概念,以获得最优增强因子。实验结果表明,该模型有效地提高了We b对象属性标注正确率。     

Automatic intersection and traffic rule detection by mining motor-vehicle GPS trajectories

The generation of road networks from ubiquitous motor-vehicle GPS trajectories has recently gained wide interest. However, few attempts have been made to automatically extract road network properties such as intersections and traffic rules to facilitate the production of high-quality routable maps. For urban street networks, the vehicle trajectory logged by a GPS receiver tends to be straight on streets and curved at intersections although the local deviation exists due to vehicle paths deviating from road centrelines and GPS positioning errors. This paper uses large curved trajectories at traffic intersections and presents novel algorithms for automatically detecting road intersections and traffic rules. Two inherent issues related to GPS trajectories have been resolved using the proposed approach. First, the serious fluctuations of vehicle trajectories due to multipath reflectivity from high-rise buildings have been eliminated, thereby enabling the effective detection of real curved trajectories occurring at traffic intersections. Second, the heterogeneity of traffic density has been considered when using the curved trajectories to automatically detect road intersections. The proposed algorithm was implemented using open-source software libraries and tested using large taxi trajectories collected in Suzhou City, China. A total of 285 at-grade intersections were detected automatically, and dynamic traffic rules were elucidated for each intersection. Compared with the manually interpreted results, the detection results were high quality and provided detailed information for the construction of a routable map.     

A machine learning approach to predict production time using real-time RFID data in industrialized building construction

Industrialized building construction is an approach that integrates manufacturing techniques into construction projects to achieve improved quality, shortened project duration, and enhanced schedule predictability. Time savings result from concurrently carrying out factory operations and site preparation activities. In an industrialized building construction factory, the accurate prediction of production cycle time is crucial to reap the advantage of improved schedule predictability leading to enhanced production planning and control. With the large amount of data being generated as part of the daily operations within such a factory, the present study proposes a machine learning approach to accurately estimate production time using (1) the physical characteristics of building components, (2) the real-time tracking data gathered using a radio frequency identification system, and (3) a set of engineered features constructed to capture the real-time loading conditions of the job shop. The results show a mean absolute percentage error and correlation coefficient of 11% and 0.80, respectively, between the actual and predicted values when using random forest models. The results confirm the significant effects of including shop utilization features in model training and suggest that predicting production time can be reasonably achieved.     

Automatic classification of ambulatory movements and evaluation of energy consumptions utilizing accelerometers and a barometer

This paper describes a method to evaluate daily physical activity by means of a portable device that determines the type of physical activity based on accelerometers and a barometer. Energy consumption of a given type of physical activity was calculated according to relative metabolic ratio (RMR) of each physical activity type that reflects exercise intensity of activities. Special attention was paid to classification algorithms for activity typing that identify detailed ambulatory movements considering vertical movements, such as stair/slope climbing or use of elevators. A portable measurement device with accelerometers and a barometer, and a Kalman filter was designed to detect the features of vertical movements. Furthermore, walking speed was calculated by an equation which estimates the walking speed as a function of signal energy of vertical body acceleration during walking. To confirm the usefulness of the method, preliminary experiments were performed with healthy young and elderly subjects. The portable device was attached to the waist. A standard accelerometer based calorie counter was also attached for comparison. Experimental results showed that the proposed method feasibly classified the type of ambulatory physical activities; level walking, stair going up and down and elevator use. It was suggested that the consideration of vertical movements made a significant improvement in the estimation of energy consumptions, and the proposed method provides better estimation of physical activity compared to the conventional calorie counter.     

Automated ergonomic risk monitoring using body-mounted sensors and machine learning

Workers in various industries are often subject to challenging physical motions that may lead to work-related musculoskeletal disorders (WMSDs). To prevent WMSDs, health and safety organizations have established rules and guidelines that regulate duration and frequency of labor-intensive activities. In this paper, a methodology is introduced to unobtrusively evaluate the ergonomic risk levels caused by overexertion. This is achieved by collecting time-stamped motion data from body-mounted smartphones (i.e., accelerometer, linear accelerometer, and gyroscope signals), automatically detecting workers’ activities through a classification framework, and estimating activity duration and frequency information. This study also investigates various data acquisition and processing settings (e.g., smartphone’s position, calibration, window size, and feature types) through a leave-one-subject-out cross-validation framework. Results indicate that signals collected from arm-mounted smartphone device, when calibrated, can yield accuracy up to 90.2% in the considered 3-class classification task. Further post-processing the output of activity classification yields very accurate estimation of the corresponding ergonomic risk levels. This work contributes to the body of knowledge by expanding the current state in workplace health assessment by designing and testing ubiquitous wearable technology to improve the timeliness and quality of ergonomic-related data collection and analysis.     

Exploring the synergetic effects of sample types on the performance of ensembles for credit risk and corporate bankruptcy prediction

Credit risk and corporate bankruptcy prediction has widely been studied as a binary classification problem using both advanced statistical and machine learning models. Ensembles of classifiers have demonstrated their effectiveness for various applications in finance using data sets that are often characterized by imperfections such as irrelevant features, skewed classes, data set shift, and missing and noisy data. However, there are other corruptions in the data that might hinder the prediction performance mainly on the default or bankrupt (positive) cases, where the misclassification costs are typically much higher than those associated to the non-default or non-bankrupt (negative) class. Here we characterize the complexity of 14 real-life financial databases based on the different types of positive samples. The objective is to gain some insight into the potential links between the performance of classifier ensembles (BAGGING, AdaBoost, random subspace, DECORATE, rotation forest, random forest, and stochastic gradient boosting) and the positive sample types. Experimental results reveal that the performance of the ensembles indeed depends on the prevalent type of positive samples.