Direct induction of compact rule-based classifiers for resource mapping

Abstract

Rule-based classifiers are used regularly with geographical information systems to map categorical attributes on the basis of a set of numeric or unordered categorical attributes. Although a variety of methods exist for inducing rule-based classifiers from training data, these tend to produce large numbers of rules when the data has noise. This paper describes a method for inducing compact rule-sets whose classification accuracy can, at least in some domains, compare favourably with that achieved by larger less succinct rule-sets produced by alternative methods. One rule is induced for each output class. The condition list for this rule represents a box in n-dimensional attribute space, formed by intersecting conditions which exclude other classes. Despite this simplicity, the classifier performed well in the test application prediction of soil classes in the Port Hills, New Zealand, on the basis of regolith type and topographic attributes obtained from a digital terrain model.     

A hybrid ensemble learning method for tourist route recommendations based on geo-tagged social networks

Geo-tagged travel photos on social networks often contain location data such as points of interest (POIs), and also users’ travel preferences. In this paper, we propose a hybrid ensemble learning method, BAyes-Knn, that predicts personalized tourist routes for travelers by mining their geographical preferences from these location-tagged data. Our method trains two types of base classifiers to jointly predict the next travel destination: (1) The K-nearest neighbor (KNN) classifier quantifies users’ location history, weather condition, temperature and seasonality and uses a feature-weighted distance model to predict a user’s personalized interests in an unvisited location. (2) A Bayes classifier introduces a smooth kernel function to estimate a-priori probabilities of features and then combines these probabilities to predict a user’s latent interests in a location. All the outcomes from these subclassifiers are merged into one final prediction result by using the Borda count voting method. We evaluated our method on geo-tagged Flickr photos and Beijing weather data collected from 1 January 2005 to 1 July 2016. The results demonstrated that our ensemble approach outperformed 12 other baseline models. In addition, the results showed that our framework has better prediction accuracy than do context-aware significant travel-sequence-patterns recommendations and frequent travel-sequence patterns.     

A spatial heterogeneity-based rough set extension for spatial data

When classical rough set (CRS) theory is used to analyze spatial data, there is an underlying assumption that objects in the universe are completely randomly distributed over space. However, this assumption conflicts with the actual situation of spatial data. Generally, spatial heterogeneity and spatial autocorrelation are two important characteristics of spatial data. These two characteristics are important information sources for improving the modeling accuracy of spatial data. This paper extends CRS theory by introducing spatial heterogeneity and spatial autocorrelation. This new extension adds spatial adjacency information into the information table. Many fundamental concepts in CRS theory, such as the indiscernibility relation, equivalent classes, and lower and upper approximations, are improved by adding spatial adjacency information into these concepts. Based on these fundamental concepts, a new reduct and an improved rule matching method are proposed. The new reduct incorporates spatial heterogeneity in selecting the feature subset which can preserve the local discriminant power of all features, and the new rule matching method uses spatial autocorrelation to improve the classification ability of rough set-based classifiers. Experimental results show that the proposed extension significantly increased classification or segmentation accuracy, and the spatial reduct required much less time than classical reduct.     

Using geovisual analytics to compare the performance of geographically weighted discriminant analysis versus its global counterpart,linear discriminant analysis

Geographically weighted spatial statistical methods are a family of spatial statistical methods developed to address the presence of non-stationarity in geographical processes, the so-called spatial heterogeneity. While these methods have recently become popular for analysis of spatial data, one of their characteristics is that they produce outputs that in themselves form complex multi-dimensional spatial data sets. Interpretation of these outputs is therefore not easy, but is of high importance, since spatial and non-spatial patterns in the results of these methods contain clues to causes of underlying non-stationarity. In this article, we focus on one of the geographically weighted methods, the geographically weighted discriminant analysis (GWDA), which is a method for prediction and analysis of categorical spatial data. It is an extension of linear discriminant analysis (LDA) that allows the relationship between the predictor variables and the categories to vary spatially. This produces a very complex data set of GWDA results, which include on top of the already complex discriminant analysis outputs (e.g. classifications and posterior probabilities) also spatially varying outputs (e.g. classification function parameters). In this article, we suggest using geovisual analytics to visualise results from LDA and GWDA to facilitate comparison between the global and local method results. For this, we develop a bespoke visual methodology that allows us to examine the performance of global and local classification method in terms of quality of classification. Furthermore, we are also interested in identifying the presence (or absence) of non-stationarity through comparison of the outputs of both methods. We do this in two ways. First, we visually explore spatial autocorrelation in both LDA and GWDA misclassifications. Second, we focus on relationships between the classification result and the independent variables and how they vary over space. We describe our visual analytic system for exploration of LDA and GWDA outputs and demonstrate our approach on a case study using a data set linking election results with a selection of socio-economic variables.     

Local entropy map: a nonparametric approach to detecting spatially varying multivariate relationships

The relationship between two or more variables may change over the geographic space. The change can be in parameter values (e.g., regression coefficients) or even in relation forms (e.g., linear, quadratic, or exponential). Existing local spatial analysis methods often assume a relationship form (e.g., a linear regression model) for all regions and focus only on the change in parameter values. Therefore, they may not be able to discover local relationships of different forms simultaneously. This research proposes a nonparametric approach, a local entropy map, which does not assume a prior relationship form and can detect the existence of multivariate relationships regardless of their forms. The local entropy map calculates an approximation of the Rényi entropy for the multivariate data in each local region (in the geographic space). Each local entropy value is then converted to a p-value by comparing to a distribution of permutation entropy values for the same region. All p-values (one for each local region) are processed by several statistical tests to control the multiple-testing problem. Finally, the testing results are mapped and allow analysts to locate and interactively examine significant local relationships. The method is evaluated with a series of synthetic data sets and a real data set.     

Spatial autocorrelation and data uncertainty in the American Community Survey: a critique

ABSTRACT

We argue that the use of American Community Survey (ACS) data in spatial autocorrelation statistics without considering error margins is critically problematic. Public health and geographical research has been slow to recognize high data uncertainty of ACS estimates, even though ACS data are widely accepted data sources in neighborhood health studies and health policies. Detecting spatial autocorrelation patterns of health indicators on ACS data can be distorted to the point that scholars may have difficulty in perceiving the true pattern. We examine the statistical properties of spatial autocorrelation statistics of areal incidence rates based on ACS data. In a case study of teen birth rates in Mecklenburg County, North Carolina, in 2010, Global and Local Moran’s I statistics estimated on 5-year ACS estimates (2006–2010) are compared to ground truth rate estimates on actual counts of births certificate records and decennial-census data (2010). Detected spatial autocorrelation patterns are found to be significantly different between the two data sources so that actual spatial structures are misrepresented. We warn of the possibility of misjudgment of the reality and of policy failure and argue for new spatially explicit methods that mitigate the biasedness of statistical estimations imposed by the uncertainty of ACS data.     

Simulating visit probability distributions within planar space-time prisms

The space-time prism is key concept in time geography and moving objects databases; it demarcates all locations that a mobile object can occupy given anchor locations and times and a maximum velocity for travel. Although the prism’s spatial and temporal extent is widely applied as a measure of accessibility and object locational uncertainty, until recently little attention has been paid to the properties of the prism interior such as the probabilities of the object visiting different locations within the prism. Better understanding of the visit probability distribution within the prism can improve theoretical understanding as well as refine the prism as a practical measure of space-time accessibility and object uncertainty. This paper presents two methods for modeling the distribution of visit probabilities within planar space-time prisms: (1) a directed Random Walk method for discrete space and time, and (2) a truncated Brownian Bridges method for continuous space and time. We illustrate these methods and demonstrate the effect of prism and mobility parameters on the visit probability distributions within the prism.     

贝叶斯专家系统分类器中专家知识的自动提取

专家知识提取是长期以来遥感专家系统分类器应用过程中存在的瓶颈问题,该文重点解决如何实现贝叶斯专家系统分类器中专家知识自动提取和知识库建立的问题。基于参考样点统计分析得出的规律,提出专家知识与参考信息间关系的假设,并设计专家知识自动提取方法,即分类类型先验概率和条件概率估计方法。为了验证专家知识自动提取方法的准确性和有效性,应用模拟数据和实际研究区域数据进行分类并评价其精度。结果表明,基于参考样点统计分析能够获得较高精度的各类型分布先验概率和条件概率,从而实现专家知识自动提取,有效解决现有贝叶斯专家系统分类器中存在的瓶颈问题。     

Same space – different perspectives: comparative analysis of geographic context through sketch maps and spatial video geonarratives

ABSTRACT

The importance of including a contextual underpinning to the spatial analysis of social data is gaining traction in the spatial science community. The challenge, though, is how to capture these data in a rigorous manner that is translational. One method that has shown promise in achieving this aim is the spatial video geonarrative (SVG), and in this paper we pose questions that advance the science of geonarratives through a case study of criminal ex-offenders. Eleven ex-offenders provided sketch maps and SVGs identifying high-crime areas of their community. Wordmapper software was used to map and classify the SVG content; its spatial filter extension was used for hot spot mapping with statistical significance tested using Monte Carlo simulations. Then, each subject’s sketch map and SVG were compared. Results reveal that SVGs consistently produce finer spatial-scale data and more locations of relevance than the sketch maps. SVGs also provide explanation of spatial-temporal processes and causal mechanisms linked to specific places, which are not evident in the sketch maps. SVG can be a rigorous translational method for collecting data on the geographic context of many phenomena. Therefore, this paper makes an important advance in understanding how environmentally immersive methods contribute to the understanding of geographic context.     

Flow Cross K-function: a bivariate flow analytical method

Spatial flow data represent meaningful interaction activities between pairs of corresponding locations, such as daily commuting, animal migration, and merchandise shipping. Despite recent advances in flow data analytics, there is a lack of literature on detecting bivariate or multivariate spatial flow patterns. In this paper we introduce a new spatial statistical method called Flow Cross K-function, which combines the Cross K-function that detects marked point patterns and the Flow K-function that detects univariate flow clustering patterns. Flow Cross K-function specifically assesses spatial dependence of two types of flow events, in other words, whether one type of flows is spatially associated with the other, and if so, whether this is according to a clustering or dispersion trend. Both a global version and a local version of Flow Cross K-function are developed. The former measures the overall bivariate flow patterns in the study area, while the latter can identify anomalies at local scales that may not follow the global trend. We test our method with carefully designed synthetic data that simulate the extreme situations. We exemplify the usefulness of this method with an empirical study that examines the distributions of taxi trip flows in New York City.     

Guest editorial

The past decade has witnessed extensive development of measures that examine characteristics of spatial subsets (local spaces) defined with respect to a complete data set (global space). Such procedures have evolved independently in fields such as geography, GIS, cartography, remote sensing, and landscape ecology. Collectively, we label these procedures as local spatial methods. We focus on those methods that share a common goal of identifying subsets whose characteristics are statistically ‘significant’ in some way. We propose the concept of local spatial statistical analysis (LoSSA) both as an integrative structure for existing methods and as a framework that facilitates the development of new local and global statistics. By formalizing what is involved when a particular local statistic is used, LoSSA helps to reveal the key features and limitations of the procedure. These include a consideration of the nature of the spatial subsets, their spatial relationship to the complete data set, and the relationship between a given global statistic and the corresponding local statistics computed for the data set.     

基于地理实体的面向对象矢量模型设计

空间数据模型是GIS实现地理空间数据组织、表达、分析、处理和应用的基础。针对基于要素的地理实体矢量描述方法的不足,从地理空间信息一体化管理角度,提出面向对象矢量模型(OOVM),探讨模型体系结构、空间对象的描述方法和空间数据管理模式。该模型利用面向对象的技术,将地理实体抽象为不同的空间对象,并将各空间对象的标识符、属性与方法封装在一起,便于网络环境下地理空间信息的存储和分布式管理。以公司信息管理为例,设计了基于OOVM的分布式空间数据组织过程,为空间数据库的建设和数据共享提供新的方法和思路。     

Extracting urban landmarks from geographical datasets using a random forests classifier

ABSTRACT

Urban landmarks are of significant importance to spatial cognition and route navigation. However, the current landmark extraction methods mainly focus on the visual salience of landmarks and are insufficient for obtaining high extraction accuracy when the size of the geographical dataset varies. This study introduces a random forests (RF) classifier combining with the synthetic minority oversampling technique (SMOTE) in urban landmark extraction. Both GIS and social sensing data are employed to quantify the structural and cognitive salience of the examined urban features, which are available from basic spatial databases or mainstream web service application programming interfaces (APIs). The results show that the SMOTE-RF model performs well in urban landmark extraction, with the values of recall, precision, F-measure and AUC reaching 0.851, 0.831, 0.841 and 0.841, respectively. Additionally, this method is suitable for both large and small geographical datasets. The ranking of variable importance given by this model further indicates that certain cognitive measures – such as feature class, Weibo popularity and Bing popularity – can serve as crucial factors for determining a landmark. The optimal variable combination for landmark extraction is also acquired, which might provide support for eliminating the variable selection requirement in other landmark extraction methods.     

基于站点观测数据的气温空间化方法评述

基于统计学的插值方法是地理学、生态学领域研究气温空间化的主要方法之一,对获取精细化气温数据进行生态模拟具有重要意义。结合国内外气温空间插值的主要研究成果,对常用气温空间化方法进行了归纳、对比,探讨各种方法的适用性和不足之处,从而为涉及气温空间化的具体研究提供一定的参考,并探讨了各类方法优化的方向。不同方法的对比分析结果表明:各种气温空间化方法各有所长,在具体的应用中都取得过较好的效果,但并不存在普适性的方法,在实际应用时必须针对研究区域具体的地理特征进行方法适用性验证或对各类方法中的具体参数进行改进,才能实现区域气温的空间最优化模拟。根据气温场的物理分布特征,结合GIS技术,考虑地形等更多的相关因子以提高气温分布微观细节的模拟精度是未来重要的发展趋势。     

国土空间规划:人口和城乡布局单幅总图的研制

研制人口与城乡布局是编制空间规划的必然要求和重点任务之一,旨在促进人口及城乡布局与资源环境承载能力相匹配,形成与“三区三线”相适应的人口和城乡布局,推进人口和城乡可持续发展。本研究提出一套“自上而下”与“自下而上”集成的人口与城乡布局研制方法。首先,总量控制,采用队列要素法、联合国法等完成基于行政区划单元的人口与城镇化水平预测;其次,因地制宜,以栅格为基本评价单元,实现城乡人口空间化,根据国土空间规划“三区三线”底图,核算现状超载和新增承载人口;最后,弹性集成人口与城镇化水平空间集疏态势以及地方国民经济发展诉求,划分人口增长地域类型、城镇增长类型、城镇规模等级等,完成人口与城乡布局。在此基础上,本研究以省级空间规划试点福建省为案例,将研制方法与研制实践相结合,实现福建省空间规划的人口与城乡布局总图绘制。以期为各尺度国土空间规划中的人口与城乡布局研制提供参考。     

空间数据统计分析的思想起源与应用演化

系统总结了空间数据统计分析的发展历程,并分为五个时期：① 早期孕育（计量革命之前）,其重要思想是19世纪初德国的区位论;② 计量革命（1950-1960年代）,主要是经典统计学的应用和理论探索;③ 空间统计学（1970-1980年代）,重点是空间点数据、面数据和空间连续性数据的分析;④ 成熟与扩散（1990-2000年代）,空间数据统计分析发展成熟并快速向其他领域扩散;⑤ 时空大数据（2010年以后）。换句话说,计量革命开始后的空间数据统计分析大约每20年有重要的新技术或方法出现,到现在已经具有成熟、系统化的方法和显著的社会效益。而在当前的时空大数据时期,其发展需要计算机科学家、统计学家和地理学家等不同学科领域人员的共同努力。     

Combining spatial transition probabilities for stochastic simulation of categorical fields

Categorical spatial data, such as land use classes and socioeconomic statistics data, are important data sources in geographical information science (GIS). The investigation of spatial patterns implied in these data can benefit many aspects of GIS research, such as classification of spatial data, spatial data mining, and spatial uncertainty modeling. However, the discrete nature of categorical data limits the application of traditional kriging methods widely used in Gaussian random fields. In this article, we present a new probabilistic method for modeling the posterior probability of class occurrence at any target location in space-given known class labels at source data locations within a neighborhood around that prediction location. In the proposed method, transition probabilities rather than indicator covariances or variograms are used as measures of spatial structure and the conditional or posterior (multi-point) probability is approximated by a weighted combination of preposterior (two-point) transition probabilities, while accounting for spatial interdependencies often ignored by existing approaches. In addition, the connections of the proposed method with probabilistic graphical models (Bayesian networks) and weights of evidence method are also discussed. The advantages of this new proposed approach are analyzed and highlighted through a case study involving the generation of spatial patterns via sequential indicator simulation.     

基于GeoDA的哈大齐工业走廊GDP空间关联性

利用地理空间分析方法和空间分析软件,分析了哈大齐工业走廊2008年各县GDP空间分布状况,包括各县之间GDP水平的空间关联性、各县GDP水平与人口的空间关联性.研究发现:该时期内哈大齐工业走廊各县GDP水平在空间关联性上成正相关,相关系数较小,GDP空间分布并非表现出完全随机性,而是表现出空间相似值之间的空间聚集.GD...     

我国省域创新产出的空间特征和时空演化——基于探索性空间数据分析的实证

以我国31个省域作为空间观测单元,以专利申请受理数作为创新产出的衡量指标,对我国1997～2008年期间省域创新产出的空间分布进行了探索性空间数据分析(ESDA)。通过计算区位基尼系数和集中度指数,发现我国的创新活动显示了相当高水平的空间集中,并且这种集中程度在过去的十多年里表现出了稳定的增长趋势;对全局的Moran’s I统计分析表明:省际创新活动之间存在着显著的空间自相关(空间依赖性),证明了知识溢出的存在性和空间局限性;对局部的Moran’s I分析进一步揭示了省际创新活动水平的相关模式,Moran散点图刻画了创新活动的空间集聚模式及其时空演变态势。研究结果说明经过十几年的发展,我国省域创新活动的地域性特征十分显著。     

Spatial interpolation using conditional generative adversarial neural networks

ABSTRACT

Spatial interpolation is a traditional geostatistical operation that aims at predicting the attribute values of unobserved locations given a sample of data defined on point supports. However, the continuity and heterogeneity underlying spatial data are too complex to be approximated by classic statistical models. Deep learning models, especially the idea of conditional generative adversarial networks (CGANs), provide us with a perspective for formalizing spatial interpolation as a conditional generative task. In this article, we design a novel deep learning architecture named conditional encoder-decoder generative adversarial neural networks (CEDGANs) for spatial interpolation, therein combining the encoder-decoder structure with adversarial learning to capture deep representations of sampled spatial data and their interactions with local structural patterns. A case study on elevations in China demonstrates the ability of our model to achieve outstanding interpolation results compared to benchmark methods. Further experiments uncover the learned spatial knowledge in the model’s hidden layers and test the potential to generalize our adversarial interpolation idea across domains. This work is an endeavor to investigate deep spatial knowledge using artificial intelligence. The proposed model can benefit practical scenarios and enlighten future research in various geographical applications related to spatial prediction.