Search

Human Learning Optimization (HLO) is an emerging meta-heuristic with promising potential. Although HLO can be directly applied to real-coded problems as a binary algorithm, the search efficiency may be significantly spoiled due to “the curse of dimensionality”. To extend HLO, Continuous HLO (CHLO) is developed to solve real-values problems. However, the research on CHLO is still in its initial stages, and further efforts are needed to exploit the effectiveness of the CHLO. Therefore, this paper proposes a novel continuous human learning optimization with enhanced exploitation (CHLOEE), in which the social learning operator is redesigned to perform global search more efficiently so that the individual learning operator is relieved to focus on performing local search for enhancing the exploitation ability. Finally, the CHLOEE is evaluated on the benchmark problem and compared with CHLO as well as recent state-of-the-art meta-heuristics. The experimental results show that the proposed CHLOEE has better optimization performance.

This paper considers the problem of designing a questionnaire for categorization of subjects. In order to well-organize the categorical structures of subjects, co-cluster information is utilized based on two different co-clustering approaches under multi-view situations. A comparative experiment studies the characteristics of fuzzy co-clustering-based categorization, where all categories are simultaneously extracted in a batch process, and possibilistic partition-based categorization, where each co-cluster is independently extracted under each view of categories.

Categorization of subjects is a basic approach for summarizing the results of various questionnaires. Co-clustering is a technique for simultaneous co-clustering of mutually familiar objects and items such that each co-cluster is formed by the subject group with their typical questions. This research considers such a situation that a questionnaire is designed for finding some pre-defined categories, which are characterized by some typical questions, but some questions may not be necessarily fit to the target categories. Then, fuzzy co-clustering is performed in conjunction with evaluation of the responsibility of each question for the target categorization. The proposed fuzzy memberships are constructed under a hierarchical structure of category characterization and question evaluation. The characteristic feature of the proposed method is demonstrated through numerical experiments with an artificial data set.

Currently, most tropical cyclone (TC) rapid intensification (RI) prediction studies are conducted based on a subset of the SHIPS database using a relatively simple model structure. However, variables (features) in the SHIPS database are built upon human expertise in TC intensity studies based on hard and subjective thresholds, and they should be explored thoroughly to make full use of the expertise. Based on the complete SHIPS data, this study constructs a complicated artificial intelligence (AI) system that handles feature engineering and selection, imbalance, prediction, and hyper parameter-tuning, simultaneously. The complicated AI system is used to further improve the performance of the current studies in RI prediction, and to identify other essential SHIPS variables that are ignored by previous studies with variable importance scores. The results outperform most of the earlier studies by approximately 21–50% on POD (Probability Of Detection) with reduced FAR (False Alarm Rate). This study built a baseline for future work on new predictor identification with more complicated AI techniques.

In hopes of better understanding the rapid intensification (RI) of tropical cyclones, the classification technique as a data mining process is used in this mining experiment. The mining results are expected to increase accurate forecasting abilities for RI through exhaustive data distillation. In this work, the Statistical Hurricane Intensity Prediction Scheme (SHIPS) database for the Atlantic basin during the period 1982–2009 is used as the data source and the Waikato Environment for Knowledge Analysis (WEKA) software is used for various classifier implementations. As in most classification applications, accuracies in model building with training data may be high. However, accuracies with testing data usually deteriorate. Various special steps are carried out in an effort to improve the accuracy. These steps include setting the cost parameters for overcoming the unbalanced RI samples, temporal averages of variable values for more accurate environmental estimation, feature filtering for irrelevant feature removal, and subset feature selections. The best performance measures of the training results are above 90% for probability of detection (POD) with 10%–20% false alarm ratios (FARs) for cases of RI within 24 h. However, the performance on the testing data is not as good. The reported RI forecasting accuracies in this work are lower than the goals set by NOAA in their Hurricane Forecast Improvement Project. Nevertheless, this work sheds light on the future direction of RI investigations using data mining techniques. Many more studies are needed before we can fully understand the potential and/or limitations of data mining techniques in RI investigations.

Pollution is one of the main negative outcomes for rapid economic growth without sustainable development in China. Different types of pollutions are harming people's health and the impacts of pollution on environment and people's health could last for decades. Fine particulate matter(PM2.5), which is one of most common types of air pollutions in China, could penetrate and sediment in human's respiratory system and cause different kind of respiratory diseases. Research has shown the strong association between Aerosol Optical Depth (AOD) and PM2.5. For this reason, remote sensing imagery could be used to estimate the level of PM2.5 concentration near ground. With utilizing PM2.5 dataset estimated by Socioeconomic Data and Applications Center (SEDAC) and machine learning approach, this paper is aimed to provide spatiotemporal comparison of PM2.5 concentrations in China. Result from this analysis could help people to better understand the recent history and current status of PM2.5 pollution in China.

Given the potentially high impact of accurate financial market forecasting, there has been considerable research on time series analysis for financial markets. We present a new Intelligent Hybrid Weighted Fuzzy (IHWF) time series model to improve forecasting accuracy in financial markets, which are complex nonlinear time-sensitive systems, influenced by many factors. The IHWF model uniquely combines Empirical Mode Decomposition (EMD) with a novel weighted fuzzy time series method. The model is enhanced by an Adaptive Sine-Cosine Human Learning Optimization (ASCHLO) algorithm to help find optimal parameters that further improve forecasting performance. EMD is a time series processing technique to extract the possible modes of various kinds of institutional and individual investors and traders, embedded in a given time series. Subsequently, the proposed weighted fuzzy time series method with chronological order based frequency and Neighborhood Volatility Direction (NVD) is analyzed and integrated with ASCHLO to determine the effective universe discourse, intervals and weights. In order to evaluate the performance of proposed model, we evaluate actual trading data of Taiwan Capitalization Weighted Stock Index (TAIEX) from 1990 to 2004 and the findings are compared with other well-known forecasting models. The results show that the proposed method outperforms the listing models in terms of accuracy.

Online reviews often contain user’s specific opinions on aspects (features) of items. These opinions are very useful to merchants and customers, but manually extracting them is time-consuming. Several topic models have been proposed to simultaneously extract item aspects and user’s opinions on the aspects, as well as to detect sentiment associated with the opinions. However, existing models tend to find poor aspect-opinion associations when limited examples of the required word co-occurrences are available in corpus. These models often also assign incorrect sentiment to words. In this paper, we propose a Latent embedding structured Opinion mining Topic model, called the LOT, which can simultaneously discover relevant aspect-level specific opinions from small or large numbers of reviews and to assign accurate sentiment to words. Experimental results for topic coherence, document sentiment classification, and a human evaluation all show that our proposed model achieves significant improvements over several state-of-the-art baselines.

Topic modeling has been widely used for extracting the major topics from a corpus. Each discovered topic contains a set of related individual words that describe the topic itself. The discovered topics summarize the major themes of the corpus. Recently, a few phrase-based topic models have been proposed, which simultaneously model phrases and topics. The topics discovered by these models consist of phrases besides individual words, as phrases are typically more meaningful. However, these models typically require large amounts of data to provide reliable statistics for phrase-based topic modeling, thus limiting their performance in scenarios with limited data. To address this limitation, we propose a knowledge-based topic model that incorporates two types of pre-identified external knowledge for topical phrase discovery: Phrase knowledge, and phrase correlation knowledge. Phrase knowledge guides the discovery of meaningful phrases by leveraging a set of pre-identified exemplary phrases; Phrase correlation knowledge guides the discovery of meaningful topics by exploiting a set of pre-identified pairs of related phrases. Experimental results show that our method outperforms the state-of-the-art baseline on both small and large datasets, extracting more meaningful phrases and coherent topics.

This paper presents a novel meta-heuristic optimization algorithm, named Adaptive Simplified Human Learning Optimization (ASHLO), which is inspired by the human learning mechanisms. Three learning operators, i.e. the random learning operator, the individual learning operator, and the social learning operator, are developed to generate new solutions and search for the optima by mimicking the learning behaviors of humans. The numerical functions, deceptive functions and 0-1 knapsack problems are adopted as benchmark problems to validate the performance of ASHLO, and the results are compared with those of binary particle swarm optimization (BPSO), modified binary differential evolution (MBDE), the binary fruit fly optimization algorithm (bFOA) and adaptive binary harmony search (ABHS). The experimental results demonstrate that the developed ASHLO significantly outperforms BPSO, MBDE, bFOA and ABHS and has a robust search ability for various problems. With the adaptive strategy, the search ability of ASHLO is improved further especially on the high-dimensional and complicated problems. Considering the ease of implementation, the excellence of global search ability and the robustness for various problems, ASHLO is a promising optimization tool for scientific research and engineering applications.

A novel meta-heuristic named human learning optimization (HLO) is presented.Four learning operators inspired by the human learning process are developed.HLO is applied to solve multi-dimensional knapsack problems.The experimental results show that HLO is a promising optimization tool. Inspired by human learning mechanisms, a novel meta-heuristic algorithm named human learning optimization (HLO) is presented in this paper in which the individual learning operator, social learning operator, random exploration learning operator and re-learning operator are developed to generate new solutions and search for the optima by mimicking the human learning process. Then HLO is applied to solve the well-known 5.100 and 10.100 multi-dimensional knapsack problems from the OR-library and the performance of HLO is compared with that of other meta-heuristics collected from the recent literature. The experimental results show that the presented HLO achieves the best performance in comparison with other meta-heuristics, which demonstrates that HLO is a promising optimization tool.

In many operations, the US government depends on geospatial data to obtain insights and drive policy making. Federal data are usually pulled from multiple sources, aggregated differently, and formatted in an assortment of standards. Analysts have been overwhelmed by the ample amounts of inconsistent and unstructured data available to them; the main quandary is that they are unable to draw clear conclusions from that data without proper analytics. Moreover, when dealing with geospatial data, the challenge of data management can exacerbate. Geospatial data (including longitudes, latitudes, country codes, addresses and many other forms) are applied at many governmental use-cases such as: census, military operations, natural disasters, economical research, agricultural commodities' production, US trade, and many others. To address such dire needs, this paper introduces a novel method (called G-Fed) for geospatial data management at the federal government. G-Fed provides tools for streaming, validating, aggregating, and presenting data. Furthermore, G-Fed utilizes the power of big data analytics to drive its processes. Software engineering best practices are also applied within G-Fed; such as multithreading, security roles, on-demand data tools, standardization, and accessibility. The United States Government Accountability Office (GAO) published a report to point to the matter of federal data management. GAO's report identified fourteen issues that governmental agencies need to solve to overcome their data challenges. Efficiency pressures and federal budget constraints are pushing data implementations towards more automation, extra validation, and higher security. Additionally, the Open Data and the Big Data Initiatives (signed in 2012) have put more pressure on the backs of agencies to make their data (including geospatial) open and available to the public. However, without having state-of-the-art data repositories, it is not possible to address such demands. This paper addresses these issues by presenting a step-by-step set of guidelines. G-Fed aims to help agencies organize geospatial data in a manner that makes it possible to publish, and be provided online to the keen American public. A data repository of public government data is used for experimentation of G-Fed. Experimental results are recorded and presented.

Financial forecasting is an extremely challenging task given the complex, nonlinear nature of financial market systems. To overcome this challenge, we present an intelligent weighted fuzzy time series model for financial forecasting, which uses a sine-cosine adaptive human learning optimization (SCHLO) algorithm to search for the optimal parameters for forecasting. New weighted operators that consider frequency based chronological order and stock volume are analyzed, and SCHLO is integrated to determine the effective intervals and weighting factors. Furthermore, a novel short-term trend repair operation is developed to complement the final forecasting process. Finally, the proposed model is applied to four world major trading markets: the Dow Jones Index (DJI), the German Stock Index (DAX), the Japanese Stock Index (NIKKEI), and Taiwan Stock Index (TAIEX). Experimental results show that our model is consistently more accurate than the state-of-the-art baseline methods. The easy implementation and effective forecasting performance suggest our proposed model could be a favorable market application prospect.

Recent developments in remote sensing have made it possible to instrument and sense the physical world with high resolution and fidelity. Consequently, very large spatio-temporal environmental data sets, have become available to the research community. Such data consists of time-series, starting as early as 1973, monitoring up to thousands of environmental parameters, for each spatial region of a resolution as low as \(0.5'\times 0.5'\). To make this flood of data actionable, in this work, we employ a data driven approach to detect and predict natural hazards. Our supervised learning approach learns from labeled historic events. We describe each event by a three-mode tensor, covering space, time and environmental parameters. Due to the very large number of environmental parameters, and the possibility of latent features hidden within these parameters, we employ a tensor factorization approach to learn latent factors. As the corresponding tensors can grow very large, we propose to employ an outlier-score for sparsification, thus explicitly modeling interesting (location, time, parameter) triples only. In our experimental evaluation, we apply our data-driven learning approach to the use-case of predicting the rapid-intensification of tropical storms. Learning from past tropical storms, we show that our approach is able to predict the future rapid-intesification of tropical storms with high accuracy, matching the accuracy of domain specific solutions, yet without using any domain knowledge.

Feature selection can efficiently improve the accuracy of classification and reduce the measurement, storage and computation demands, and thus it has been applied in biomedical research increasingly. Considering the non-deterministic polynomial-time hard characteristic of feature selection, meta-heuristics are introduced into feature selection in biomedicine on account of their excellent global search ability. However, most of biomedical problems are characterized by high dimensionality, which is a challenge for feature selection methods based on meta-heuristics due to the curse of dimensionality. Thus, six meta-heuristics, that is, a genetic algorithm, particle swarm optimization, ant colony optimization, harmony search, differential evolution, and quantum-inspired evolutionary algorithm, which are widely studied in the meta-heuristic community, are introduced into feature selection in this paper and the performance of the algorithms is analysed and compared with each other for solving feature selection ...

This study applies a data mining technique called association rule mining to the analysis of intensity changes of North Atlantic tropical cyclones (TCs). The “best track” data from the National Hurricane Center and the Statistical Hurricane Intensity Prediction Scheme databases were stratified into tropical depressions, tropical storms, and category 1–5 hurricanes based on the Saffir–Simpson hurricane scale. After stratification, the seven resulting groups of TCs plus two additional aggregation groups were further separated into intensifying, weakening, and stable TCs. The analysis of the stratified data for preprocessing revealed that faster northward storm motion (the meridional component of storm motion) favors tropical storm intensification but does not favor the intensification of hurricanes. Intensifying tropical storms are more strongly associated with a higher convergence in the upper atmosphere (200-hPa relative eddy momentum flux convergence) than weakening tropical storms, while intensifying hurricanes are more strongly associated with lower convergence values. The mined association rules showed that cofactors usually display higher-intensity prediction power in the stratified TC groups. The data mining results also identified a predictor set with fewer factors but improved probabilities of rapid intensification. This study found that the data mining technique not only sheds light on the roles of multiple-associated physical processes in tropical cyclone development—especially in rapid intensification processes—but also will help improve TC intensity forecasting. This paper provides an outline on how to use this data mining technique and how to overcome low occurrences of mined conditions in order to improve TC intensity forecasting capabilities.

WebGIS (also known as web‐based GIS and Internet GIS) denotes a type of Geographic Information System (GIS), whose client is implemented in a Web browser. WebGISs have been developed and used extensively in real‐world applications. However, when such a complex web‐based system involves the dissemination of large volumes of data and/or massive user interactions, its performance can become an issue. In this paper, we first identify several major potential performance problems with WebGIS. Then, we discuss several possible techniques to improve the performance. These techniques include the use of pyramids and hash indices on the server side to handle large images. To resolve server‐side conflicts originating from concurrent massive access and user interactions, we suggest clustering and multithreading techniques. Multithreading is also used to break down the long sequential, layer‐based data access to concurrent data access on the client side. Caching is suggested as a means to enhance concurrent data access...

We present our XML-based Distributed Metadata Server (Dimes) - which comprises a flexible metadata model, search software, and a Web-based interface - to support multilevel metadata access, and introduce two prototype systems. Our Scientific Data and Information Super Server (SDISS), which is based on Dimes and GDS, solves accurate data-search and outdated data-link problems by integrating metadata with the data systems. On the implementation front, we combine independent components and open-source technologies into a coherent system to dramatically extend system capabilities. Obviously, our approach can be applied to other scientific communities, such as bioinformatics and space science.

Many of the challenges of the next century will have physical dimensions, such as tsunamis, hurricanes, and climate change as well as human dimensions such as economic crises, epidemics, and emergency responses. With pioneering editors and expert contributors, Advanced Geoinformation Science explores how certain technical aspects of geoinformation have been used and could be used to address such global issues. The editors and chapter authors have been involved in global initiatives and research problems, such as Global Earth Observation System of Systems (GEOSS), Digital Earth, Air Quality, Public Health, and Cloud Computing. The book delineates the problems communities are likely to face and how advanced geoinformation science can be a part of their solution. It introduces different methods in collecting spatial data as the initial feeds to geoinformation science and computing platforms. It discusses systems for data management, data integration and analysis, the geoinformation infrastructure, as well as knowledge capture, formatting, and utilization. The book then explores a variety of geoinformation applications, highlighting environmental, agriculture, and urban planning uses. Geoinformation science encompasses more than just traditional technologies such as Remote Sensing, GIS, GPS, and supporting disciplines. And although the science continues to become more multidisciplinary, the literature remains compartmentalized according to the traditional disciplinary boundaries. Capturing recent developments in geoinformation science and linking IT with a wide range of Earth sciences, the authors explain how advanced technology and concepts play a significant role in recent advancements. Enhanced with forty-four color illustrations, live examples such as GOS and AirNow, and insights from NASA, EPA, and USGS, the book provides a vision for the future and explores how to bring that vision into reality.

Principal Component Analysis (PCA) has been extensively used in different fields including earth science for spatial pattern identification. However, the intrinsic linear feature associated with standard PCA prevents scientists from detecting nonlinear structures. Kernel-based principal component analysis (KPCA), a recently emerging technique, provides a new approach for exploring and identifying nonlinear patterns in scientific data. In this paper, we recast KPCA in the commonly used PCA notation for earth science communities and demonstrate how to apply the KPCA technique into the analysis of earth science data sets. In such applications, a large number of principal components should be retained for studying the spatial patterns, while the variance cannot be quantitatively transferred from the feature space back into the input space. Therefore, we propose a KPCA pattern selection algorithm based on correlations with a given geophysical phenomenon. We demonstrate the algorithm with two widely used data sets in geophysical communities, namely the Normalized Difference Vegetation Index (NDVI) and the Southern Oscillation Index (SOI). The results indicate the new KPCA algorithm can reveal more significant details in spatial patterns than standard PCA.

Since its interception in the 1960s (GIS World, 1996; Tomlinson, 1998), Geographic Information System (GIS) has been well established as an information system for acquiring, storing, processing, analyzing and visualizing spatial data. The success of GIS is reflected from the following developments: (1) numerous GIS software have been developed and are widely used; (2) A large amount of spatial datasets have been collected and used; (3) Many books and papers have been written to introduce and discuss different aspects of GIS. They range from providing an overview of GIS (Davis, 2003; Tomlinson, 1998), discussing selected advanced topics (Kreveld et al., 1997; O’Sullivan and Unwin, 2002), to recording the relevant history (Foresman, 1997).

We have witnessed the accumulation of petabyes of geospatial data in the past decades, and, currently, terabytes of data are collected every day. These geospatial data are crucial in supporting decision making and emergency response. It becomes increasingly important to deliver these datasets in a timely fashion to the decision support or emergency response systems. Network GIS provides a vehicle to facilitate this delivering process. But to deliver efficiently such large volume of data and to handle large number of concurrent users, the performance of Network GIS needs to be improved to a level that different types of applications, especially near real time applications, can be satisfied. In this circumstance, we 1) review selected research on improving the performance of Network GIs; 2) provide insides on implementing the techniques; and 3) illustrate how to adopt the techniques in Network GIs. We expect the research and development reported here can be easily adopted by different users to accelerate the performance of various Network GIS software and applications, as well as to support the building of spatial data infrastructure to support the sharing of heterogeneous geospatial information.

Illustrates a distributed system that provides online searching, analysis and ordering capabilities for distributed Earth science data. The system is under development by a consortium led by George Mason University in a project called Seasonal-to-Interannual Earth Science Information Partners (SIESIP) as a part of a federation of information partners funded by NASA. The integrated system is composed of data, a database management system (DBMS), communication protocols, data analysis tools and a user interface. Through a Web-based Java GUI, users can search the DBMS for metadata information, conduct content-based searches, perform some initial analyses and issue an order for the selected data.

Addresses a reliable, feasible method to find geographical areas with constraints using hierarchical depth-first clustering. The method involves multi-level hierarchical clustering with a depth-first strategy, depending on whether the area of each cluster satisfies the given constraints. The attributes used in the hierarchical clustering are the coordinates of the grid data points. The constraints are an average value range and the minimum size of an area with a small proportion of missing data points. Convex-hull and point-in-polygon algorithms are involved in examining the constraint satisfaction. The method is implemented for an Earth science data set for vegetation studies - the Normalized Difference Vegetation Index (NVDI).

With explosively increasing volumes of remote sensing, modelling and other Earth science data available, and the popularity of the Internet, scientists are now facing challenges to publish and to find interesting data sets effectively and efficiently. Metadata has been recognized as a key technology to ease the searching and retrieval of Earth science data. In this paper, we discuss the DIMES (DIstributed MEtadata Server) prototype system. Designed to be flexible yet simple, DIMES uses XML to represent, store, retrieve and interoperate metadata in a distributed environment. DIMES accepts metadata in any well-formed XML format and thus assumes the "tree" semantics of metadata entries. Additional domain knowledge can be represented as specific links through XML's ID/IDREF mechanism. DIMES provides a number of mechanisms, including the "nearest-neighbor search", to navigate and to search metadata. Though started for the Earth science community, DIMES can be easily extended to serve scientific communities in other disciplines.

Current and next decade global Earth observing, other remote sensing and related climate analysis data collected by space and operational U.S. agencies such as NASA and NOAA, the European ESA, the Japanese NASDA and other international agency missions of India, China, Russia, etc. will reach unprecedented data volumes, exceeding many petabytes. This is rushing in a new era in Earth system science. Along with the technology and data volumes afforded by remote sensing, there has been an unprecedented increase of capabilities in distributed data systems in the last few years. The existence of the Internet and the World Wide Web afford users access to data at diverse distributed sites that would had been very difficult in the past and only available to specialists. These data can be accessed by a variety of scientists, applications experts and the general public. Yet, the unprecedented large volumes of such missions are also presenting formidable challenges to wide user access and require both higher bandwidths of future Internet systems as well as more focused, user-centered data productions. Specialized data productions are best achieved in federated data systems. The large data volumes present a challenge of access, storage and distribution. What is most important is not just the volume of the data itself but the information they contain. A data system's usefulness is related to the ease that its users can search and access products and as such obtain information on the actual content of data before proceeding to order large volumes of data sets which may or may not serve their needs. Here we explore different functionalities in distributed Earth observing data systems and associated interoperability options. As an example, we examine options in the Earth Science Information Partners Federation (ESIPs) funded by NASA to extend the usage of NASA remote sensing data holdings. Results of several interoperability options applicable to federated systems are also presented. We also examine challenges presented by the use of regional remote sensing missions such as hyperspectral imaging.

Remote sensing data as well as ground-based and model output data about the Earth system can be very large in volume. On the other hand, in order to use the data efficiently, scientists need to search for data based on not only metadata but also actual data values. To answer value range queries by scanning very large volumes of data is obviously unrealistic. This article studies a clustering technique on histograms of data values on predefined cells to index the cells. Through this index system, the so-called statistical range queries can be answered quickly and approximately together with an accuracy assessment. Examples of using this technique for Earth science data sets are given in this article.