Traditional Datasets Research Articles

Research on Virtual Reconstruction Technology of Tujia Brocade Handcrafts

The colorful Tujia brocade culture is formed through thousands of years’ inheritance, development, and creation in China. However, the influence of the Tujia brocade culture has been gradually weakened for the restrictions of regional and economic development. Also the traditional brocade handcrafts are on the verge of disappearing. Consequently, it is important and necessary to reconstruct traditional Tujia brocade skill by using digital protection technology. Lots of researchers have studied the reconstruction of cultural scenes through the virtual simulations of character movements in the field of intangible cultural heritage, and most studies focus only on limb movements or only on detailed hand movements, while few of them have combined the limb movements and detailed hand movements. According to the characteristics of Tujia brocade craftsmanship, a solution of virtual reconstruction of the traditional Tujia brocade handcrafts to solve the difficult problem for the simultaneous synthesis of limb movements and detailed hand movements was presented in this paper. The solution used Kinect (Kinect for windows v2) and Leap motion to capture the limb movements and detailed hand movements respectively. Then it established traditional process motion data set of Tujia brocade by integrating limb and detailed hand movement data. Finally, it adopted mean smoothing algorithm to manipulate the motion data and drove the character model to realize the virtual reconstruction of the traditional Tujia brocade handcrafts. Multiple simulation and system fluidity test results showed that the proposed method could solve the problem of simultaneous synthesis of limb movements and hand movements in virtual reconstruction of traditional handcrafts effectively.

Real-world plant species identification based on deep convolutional neural networks and visual attention

This paper investigates the issue of real-world identification to fulfill better species protection. We focus on plant species identification as it is a classic and hot issue. In tradition plant species identification the samples are scanned specimen and the background is simple. However, real-world species recognition is more challenging. We first systematically investigate what is realistic species recognition and the difference from tradition plant species recognition. To deal with the challenging task, an interdisciplinary collaboration is presented based on the latest advances in computer science and technology. We propose a novel framework and an effective data augmentation method for deep learning in this paper. We first crop the image in terms of visual attention before general recognition. Besides, we apply it as a data augmentation method. We call the novel data augmentation approach attention cropping (AC). Deep convolutional neural networks are trained to predict species from a large amount of data. Extensive experiments on traditional dataset and specific dataset for real-world recognition are conducted to evaluate the performance of our approach. Experiments first demonstrate that our approach achieves state-of-the-art results on different types of datasets. Besides, we also evaluate the performance of data augmentation method AC. Results show that AC provides superior performance. Compared with the precision of methods without AC, the results with AC achieve substantial improvement.

An automatic data integration approach to enrich ATKIS with the VGI of outdoor-sports data

With the availability of cheap location technologies, volunteered geographic information (VGI) has become one of the most relevant geographic data sources. In order to make more efficient use of such volunteered data, the VGI is often needed to be integrated to the corresponding ‘traditional’ datasets. This article proposes a specific case of data integration between (a) the topographic dataset of ATKIS maintained by German surveying and mapping agencies and (b) the Alpstein Outdoor-Sport Data (AOSD) primarily collected by numerous enthusiasts (volunteers) of jogging, hiking, biking (incl. road and mountain), skiing, etc. Considering that the VGI of AOSD reveals quite different LODs (level of details) to the traditional topographic dataset of ATKIS, the authors put forward a new approach to realize automatic and accurate data integrations between these two datasets, which demonstrates high performance in a number of large test areas and thereby has been successfully utilized for the real-world data production in the area of Alps, Europe. As a result, the integrated datasets have been published on the website to serve broad masses of outdoor-sports enthusiasts by a German Cooperation, viz. Outdooractive GmbH.

Evaluation of machine learning techniques with multiple remote sensing datasets in estimating monthly concentrations of ground-level PM2.5

Fine particulate matter (PM2.5) has been recognized as a key air pollutant that can influence population health risk, especially during extreme cases such as wildfires. Previous studies have applied geospatial techniques such as land use regression to map the ground-level PM2.5, while some recent studies have found that Aerosol Optical Depth (AOD) derived from satellite images and machine learning techniques may be two elements that can improve spatiotemporal prediction. However, there has been a lack of studies evaluating use of different machine learning techniques with AOD datasets for mapping PM2.5, especially in areas with high spatiotemporal variability of PM2.5.In this study, we compared the performance of eight predictive algorithms with the use of multiple remote sensing datasets, including satellite-derived AOD data, for the prediction of ground-level PM2.5 concentration. Based on the results, Cubist, random forest and eXtreme Gradient Boosting were the algorithms with better performance, while Cubist was the best (CV-RMSE = 2.64 μg/m3, CV-R2 = 0.48). Variable importance analysis indicated that the predictors with the highest contributions in modelling were monthly AOD and elevation.In conclusion, appropriate selection of machine learning algorithms can improve ground-level PM2.5 estimation, especially for areas with nonlinear relationships between PM2.5 and predictors caused by complex terrain. Satellite-derived data such as AOD and land surface temperature (LST) can also be substitutes for traditional datasets retrieved from weather stations, especially for areas with sparse and uneven distribution of stations.

Characterize Collective Lysosome Heterogeneous Dynamics in Live Cell with a Space- and Time-Resolved Method.

While studies of collective cell migration and bacteria swarming have tremendously promoted our fundamental knowledge of the complex systematic phenomena, the quantitative characterization of the collective organelles movement at subcellular level is yet to be fully explored. Here we tagged the lysosomes in live cells with fluorescent probe and imaged their spatial motion with wide field microscopy. To quantitatively characterize the collective lysosomal behavior with high spatiotemporal heterogeneity dynamics, we developed the particle collective analysis (PECAN) method based on the single particle tracking techniques. Thousands of trajectories were detected and analyzed in each single cell. The reliability was validated by comparing with traditional PIV method, simulated and experimental data sets. We show that the lysosomes in live cells move collectively with spatial heterogeneous and temporal long-term correlated dynamics. Furthermore, the continuous wavelet analysis suggested the existence of collective lysosomal oscillation in mouse neural cells. Generally, our method provides a practical workflow for characterizing the collective lysosomal motions which can benefit related areas such as organelles mediated drug delivery and cell activity profiling.

Multi-label learning with label-specific features by resolving label correlations

In multi-label learning, different labels may have their own inherent characteristics for distinguishing each other, in the meanwhile, exploiting the correlations among labels is another practical yet challenging task to improve the performance. In this work, we present a new method for the joint learning of label-specific features and label correlations. The key is the design of an optimization framework to learn the weight assignment scheme of features, and the correlations among labels are taken into account by constructing additional features at the same time. Through iteratively optimizing the two sets of unknown variables, which are referred to feature weights and label correlations-based features, label-specific features of each label are available to achieve multi-label classification. Comprehensive experiments on various multi-label data sets including two collected traditional Chinese medicine data sets reveal the advantages of our proposed algorithm.

Assessing the multi-scale predictive ability of ecosystem functional attributes for species distribution modelling.

Global environmental changes are rapidly affecting species’ distributions and habitat suitability worldwide, requiring a continuous update of biodiversity status to support effective decisions on conservation policy and management. In this regard, satellite-derived Ecosystem Functional Attributes (EFAs) offer a more integrative and quicker evaluation of ecosystem responses to environmental drivers and changes than climate and structural or compositional landscape attributes. Thus, EFAs may hold advantages as predictors in Species Distribution Models (SDMs) and for implementing multi-scale species monitoring programs. Here we describe a modelling framework to assess the predictive ability of EFAs as Essential Biodiversity Variables (EBVs) against traditional datasets (climate, land-cover) at several scales. We test the framework with a multi-scale assessment of habitat suitability for two plant species of conservation concern, both protected under the EU Habitats Directive, differing in terms of life history, range and distribution pattern (Iris boissieri and Taxus baccata). We fitted four sets of SDMs for the two test species, calibrated with: interpolated climate variables; landscape variables; EFAs; and a combination of climate and landscape variables. EFA-based models performed very well at the several scales (AUCmedian from 0.881±0.072 to 0.983±0.125), and similarly to traditional climate-based models, individually or in combination with land-cover predictors (AUCmedian from 0.882±0.059 to 0.995±0.083). Moreover, EFA-based models identified additional suitable areas and provided valuable information on functional features of habitat suitability for both test species (narrowly vs. widely distributed), for both coarse and fine scales. Our results suggest a relatively small scale-dependence of the predictive ability of satellite-derived EFAs, supporting their use as meaningful EBVs in SDMs from regional and broader scales to more local and finer scales. Since the evaluation of species’ conservation status and habitat quality should as far as possible be performed based on scalable indicators linking to meaningful processes, our framework may guide conservation managers in decision-making related to biodiversity monitoring and reporting schemes.

Multi-Exposure Motion Estimation Based on Deep Convolutional Networks

In motion estimation, illumination change is always a troublesome obstacle, which often causes severely performance reduction of optical flow computation. The essential reason is that most of estimation methods fail to formalize a unified definition in color or gradient domain for diverse environmental changes. In this paper, we propose a new solution based on deep convolutional networks to solve the key issue. Our idea is to train deep convolutional networks to represent the complex motion features under illumination change, and further predict the final optical flow fields. To this end, we construct a training dataset of multi-exposure image pairs by performing a series of non-linear adjustments in the traditional datasets of optical flow estimation. Our multi-exposure flow networks (MEFNet) model consists of three main components: low-level feature network, fusion feature network, and motion estimation network. The former two components belong to the contracting part of our model in order to extract and represent the multi-exposure motion features; the third component is the expanding part of our model in order to learn and predict the high-quality optical flow. Compared with many state-of-the-art methods, our motion estimation method can eliminate the obstacle of illumination change and yield optical flow results with competitive accuracy and time efficiency. Moreover, the good performance of our model is also demonstrated in some multi-exposure video applications, like HDR (high dynamic range) composition and flicker removal.

Satellite-based salt marsh elevation, vegetation height, and species composition mapping using the superspectral WorldView-3 imagery

ABSTRACTVery high resolution (VHR) space-borne data are needed to finely and continuously map salt marshes. The WorldView-3 (WV-3) sensor leverages one panchromatic, eight optical, and eight shortwave infrared (SWIR) bands at 0.31, 1.24, and 7.5 m pixel size, respectively. Although eight optical bands have been previously pansharpened, no attempt to use the 16-band superspectral data set at VHR (0.31 m) has been yet reviewed. Here, we propose to reliably pan-sharpen the 16 WV-3 predictors so as to model (artificial neural network, ANN) salt marsh elevation and vegetation height and classify species composition at VHR using calibration/validation handheld vegetation height, airborne lidar elevation, and drone blue-green-red (BGR) responses. Three models have been created over a megatidal bay (Beaussais Bay, Brittany, France) provided with mud flats, salt marshes, and polders. VHR-screened WV-3 bands very satisfactorily predicted salt marsh elevation and vegetation height responses (r = 0.86, R2 = 0.71, root mean square error (RMSE) = 0.33 m and r = 0.88, R2 = 0.77, RMSE = 5.72 cm, respectively). The WV-3 superspectral data set outperformed the eight-band multispectral and four-band traditional data sets to classify 15 salt marsh habitats (OA = 95.47, 82.33, and 69.27%, respectively). Adding WV-3-based salt marsh elevation and vegetation height augmented the 15-class classification of the superspectral and traditional data sets (OA = 97.60 and 77.47%, respectively), but not for the multispectral one (OA = 81.93%).

Airline customer lifetime value estimation using data analytics supported by social network information

Companies can improve their customer relationships and business performance via analytical applications such as estimation of customer lifetime value (CLV) and profitability, customer profiling and classification, customer retention and churn analyses. Customer Relationship Management (CRM) tools can now have access to relationship and interaction data of the customers, besides the traditional data sets such as billing information. While there has been a sharp increase in mining social and interaction data, integration of this information with the current data analytical models is limited. In this paper, we develop a new model for estimating the customer lifetime value in airline industry that integrates customers' social network and flight information. We first adopt a regression model for airline customers that can be used to estimate their CLVs. We then present a methodology to enhance this base model with customers' social network information to incorporate indirect contributions the customers make. We compare the performances of both models to show that our proposed method may improve the accuracy and reliability of models that make use of only flight related factors. We provide examples to potential customer analyses using our models for use by airline CRM applications.

Using spatiotemporal distribution of geocoded Twitter data to predict US county-level health indices

For more than three decades, the US has annually conducted Behavioral Risk Factor Surveillance System (BRFSS) surveys to capture health behavior and health status of its people. Though this kind of information at population level is important for local governments to identify local needs, traditional datasets take several years to collate and to become publicly available. Geocoded social media data can provide an alternative reflection of local health trends. Due to the large scale of data, such as approximately two billions of tweets in this work, aggregating the tweets at a population level is common practice. While alleviating the computational cost, the aggregation operation would result in the loss of information on the distribution of data over the population, and such information may be important for identifying the health behavior and health outcomes of the population. In this work, we propose statistical features constructed on-top of primary features to predict county-level health indices. The primary features include topics and linguistic patterns extracted from tweets with county-decoded information. In addition, tweeting behaviors, particularly tweeting time, are used as a predictor of the health indices. Apache Spark, an advanced cluster computing paradigm, was employed to efficiently process the large corpus of tweets, including geo-decoding the geotags, extracting low-level (primary) features, and computing the statistical features. The results show strong correlations between publicly available health indices and the features extracted from geospatially coded Twitter data. Statistical features gained higher correlation coefficients than did the aggregation ones, suggesting the potential and applicability of the proposed features in a wide spectrum of applications on data analytics at population levels. In addition, the prediction performance was also improved when the temporal information was employed. This demonstrates that the real-time analysis of social media data can provide timely insights into the health of populations.

Concealment Conserving the Data Mining of Groups & Individual

We present an overview of privacy preserving data mining, one of the most popular directions in the data mining research community. In the first part of the chapter, we presented approaches that have been proposed for the protection of either the sensitive data itself in the course of data mining or the sensitive data mining results, in the context of traditional (relational) datasets. Following that, in the second part of the chapter, we focused our attention on one of the most recent as well as prominent directions in privacy preserving data mining: the mining of user mobility data. Although still in its infancy, privacy preserving data mining of mobility data has attracted a lot of research attention and already counts a number of methodologies both with respect to sensitive data protection and to sensitive knowledge hiding. Finally, in the end of the chapter, we provided some roadmap along the field of privacy preserving mobility data mining as well as the area of privacy preserving data mining at large.

Urban spatiotemporal analysis using mobile phone data: Case study of medium- and large-sized Korean cities

Recent advanced information and communication technologies can provide more accurate and comprehensive information. In particular, mobile phone data provide a new data source for urban structures and mobility. It is important for urban and transportation planners to be able to use data containing valuable location information that is not easily obtainable from traditional datasets, such as expensive household surveys, extensive traffic counts, or aggregated socioeconomic statistics. This study explored the potential of using mobile phone data to characterize and compare urban activity and mobility patterns from daily and hourly mobile phone records across 10 cities in Korea. We compared the internal and external mobility of phone users, and calculated urban attractiveness and home-based trip length frequency distribution for comparisons of different sized cities. The spatiotemporal evolution of urban activity within a day was examined, and the spatiotemporal extension for each city was calculated, showing the degree of spatial dispersion of residences and other activity locations. We also identified urban activity subcenters and hotspots based on density and time persistence criteria, as well as their compactness. Policy makers can expect to see more applications using mobile phone data, and this study helps demonstrate the potential of such data. We hope that spatiotemporal activity analysis can provide a foundation for future research that will help improve urban and transportation policies.

Biopanning data bank 2018: hugging next generation phage display.

The 2018 update of the biopanning data bank (BDB) stores phage display data sequenced by Sanger sequencing and next generation sequencing technologies. In this work, we upgraded the database with more biopanning data sets and several new features, including (i) incorporation of next generation biopanning data and the unselected population where the target is not determined and the round of screening is zero; (ii) addition of sequencing information; (iii) improvement of browsing and searching systems and 3 D chemical structure viewer; (iv) integration of standalone tools for target-unrelated peptides analysis within conventional phage display and next generation phage display (NGPD) data. In the current version of BDB (released on 19 January 2018), the database houses 3291 sets of biopanning data collected from 1540 published articles, including 95 NGPD data sets and 3196 traditional biopanning data sets. The BDB database serves as an important and comprehensive resource for developing peptide ligands. Database URL: The BDB database is available at http://immunet.cn/bdb

Improving Recommendation Techniques by Deep Learning and Large Scale Graph Partitioning

Recommendation is very crucial technique for social networking sites and business organizations. It provides suggestions based on users’ personalized interest and provide users with movies, books and topics links that would be most suitable for them. It can improve user effectiveness and business revenue by approximately 30%, if analyzed in intelligent manner. Social recommendation systems for traditional datasets are already analyzed by researchers and practitioners in detail. Several researchers have improved recommendation accuracy and throughput by using various innovative approaches. Deep learning has been proven to provide significant improvements in image processing and object recognition. It is machine learning technique where hidden layers are used to improve outcome. In traditional recommendation techniques, sparsity and cold start are limitations which are due to less user-item interactions. This can be removed by using deep learning models which can improve user-item matrix entries by using feature learning. In this paper, various models are explained with their applications. Readers can identify best suitable model from these deep learning models for recommendation based on their needs and incorporate in their techniques. When these recommendation systems are deployed on large scale of data, accuracy degrades significantly. Social big graph is most suitable for large scale social data. Further improvements for recommendations are explained with the use of large scale graph partitioning. MAE (Mean Absolute Error) and RMSE (Root Mean Squared Error) are used as evaluation parameters which are used to prove better recommendation accuracy. Epinions, MovieLens and FilmTrust datasets are also shown as most commonly used datasets for recommendation purpose.

Sideritic ironstones as indicators of depositional environments in the Weald Basin (Early Cretaceous) SE England

AbstractThe Lower Cretaceous Wealden sideritic ironstones have a wide occurrence and great potential to aid the reconstruction of the depositional environments of the Weald Basin in SE England. However, mineralogical and geochemical datasets on the ironstones are scarce in the literature. Geochemical and mineralogical data on the sideritic ironstones are presented from the Wadhurst Clay Formation within the Weald Basin. The mineralogy of the ironstones was examined using a PANalytical X'Pert Pro X-ray diffractometer and PANalytical's HighScore Plus software. Elemental composition of the ironstones was measured using a PANalytical MiniPal2 ED-XRF (benchtop X-ray spectrometer). The examination of the mineralogy of the Wealden ironstones confirms the presence of early diagenetic siderites. The trace-element assemblage shows that the sideritic ironstones are chemically pure pointing to a freshwater origin. The sideritic ironstones reveal anoxic conditions and palaeo-salinity in the basin. More generally, it is suggested that the composition of the host rocks has significant controls on the composition of sideritic ironstones in sedimentary basins. This work reinforces the importance of the composition of sideritic ironstones as useful non-traditional data for understanding the depositional settings of sedimentary basins, especially when traditional datasets are not readily available or insufficient.

Regolith-geology mapping with support vector machine: A case study over weathered Ni-bearing peridotites, New Caledonia

Accurate maps of Earth’s geology, especially its regolith, are required for managing the sustainable exploration and development of mineral resources. This paper shows how airborne imaging hyperspectral data collected over weathered peridotite rocks in vegetated, mountainous terrane in New Caledonia were processed using a combination of methods to generate a regolith-geology map that could be used for more efficiently targeting Ni exploration. The image processing combined two usual methods, which are spectral feature extraction and support vector machine (SVM). This rationale being the spectral features extraction can rapidly reduce data complexity by both targeting only the diagnostic mineral absorptions and masking those pixels complicated by vegetation, cloud and deep shade. SVM is a supervised classification method able to generate an optimal non-linear classifier with these features that generalises well even with limited training data. Key minerals targeted are serpentine, which is considered as an indicator for hydrolysed peridotitic rock, and iron oxy-hydroxides (hematite and goethite), which are considered as diagnostic of laterite development. The final classified regolith map was assessed against interpreted regolith field sites, which yielded approximately 70% similarity for all unit types, as well as against a regolith-geology map interpreted using traditional datasets (not hyperspectral imagery). Importantly, the hyperspectral derived mineral map provided much greater detail enabling a more precise understanding of the regolith-geological architecture where there are exposed soils and rocks.

A new evolutionary algorithm for mining top-k discriminative patterns in high dimensional data

This paper presents an evolutionary algorithm for Discriminative Pattern (DP) mining that focuses on high dimensional data sets. DPs aims to identify the sets of characteristics that better differentiate a target group from the others (e.g. successful vs. unsuccessful medical treatments). It becomes more natural to extract information from high dimensionality data sets with the increase in the volume of data stored in the world (30GB/s only in the Internet). There are several evolutionary approaches for DP mining, but none focusing on high-dimensional data. We propose an evolutionary approach attributing features that reduce the cost of memory and processing in the context of high-dimensional data. The new algorithm thus seeks the best (top-k) patterns and hides from the user many common parameters in other evolutionary heuristics such as population size, mutation and crossover rates, and the number of evaluations. We carried out experiments with real-world high-dimensional and traditional low dimensional data. The results showed that the proposed algorithm was superior to other approaches of the literature in high-dimensional data sets and competitive in the traditional data sets.

White matter biomarkers from fast protocols using axially symmetric diffusion kurtosis imaging.

White matter tract integrity (WMTI) can characterize brain microstructure in areas with highly aligned fiber bundles. Several WMTI biomarkers have now been validated against microscopy and provided promising results in studies of brain development and aging, as well as in a number of brain disorders. Currently, WMTI is mostly used in dedicated animal studies and clinical studies of slowly progressing diseases, and has not yet emerged as a routine clinical tool. To this end, a less data intensive experimental method would be beneficial by enabling high resolution validation studies, and ease clinical applications by speeding up data acquisition compared with typical diffusion kurtosis imaging (DKI) protocols utilized as part of WMTI imaging. Here, we evaluate WMTI based on recently introduced axially symmetric DKI, which has lower data demand than conventional DKI. We compare WMTI parameters derived from conventional DKI with those calculated analytically from axially symmetric DKI. We employ numerical simulations, as well as data from fixed rat spinal cord (one sample) and in vivo human (three subjects) and rat brain (four animals). Our analysis shows that analytical WMTI based on axially symmetric DKI with sparse data sets (19 images) produces WMTI metrics that correlate strongly with estimates based on traditional DKI data sets (60 images or more). We demonstrate the preclinical potential of the proposed WMTI technique in in vivo rat brain (300μm isotropic resolution with whole brain coverage in a 1h acquisition). WMTI parameter estimates are subject to a duality leading to two solution branches dependent on a sign choice, which is currently debated. Results from both of these branches are presented and discussed throughout our analysis. The proposed fast WMTI approach may be useful for preclinical research and e.g. clinical evaluation of patients with traumatic white matter injuries or symptoms of neurovascular or neuroinflammatory disorders.

An alternative statistical method for characterizing low-frequency environments in sensitive laboratory settings

Many practitioners are familiar with the collection and expression of complex noise data as Ln statistical spectra. Given suitable instrumentation, it is easy to develop Ln statistics for low-frequency sound and vibration environments, as well. However, the traditional (instrument-generated) Ln datasets have some shortcomings when it comes to understanding some environments, especially in highly sensitive settings like research laboratories. The emergence of mass data storage and manipulation tools has provided an opportunity to employ a somewhat subtler method to characterize these environments. In this presentation, we will describe this analytical technique, which provides statistical representations similar to the Ln system. However, the technique further allows an explicit invocation of timescale, allowing the user to choose an arbitrary analytical timescale relevant to sensitive operations or expected environmental transients or other forcings. We will also show examples that demonstrate why the approach might find particular use in technical settings like nanotechnology laboratories and other environments supporting research in the physical sciences.