Julian Hagenauer

Toward mapping land-use patterns from volunteered geographic information

A large number of applications have been launched to gather geo-located information from the public. This article introduces an approach toward generating land-use patterns from volunteered geographic information (VGI) without applying remote-sensing techniques and/or engaging official data. Hence, collaboratively collected OpenStreetMap (OSM) data sets are employed to map land-use patterns in Vienna, Austria. Initially the spatial pattern of the landscape was delineated and thereafter the most relevant land type was assigned to each land parcel through a hierarchical GIS-based decision tree approach. To evaluate the proposed approach, the results are compared with the Global Monitoring for Environment and Security Urban Atlas (GMESUA) data. The results are compared in two ways: first, the texture of the resulting land-use patterns is analyzed using texture-variability analysis. Second, the attributes assigned to each land segment are evaluated. The achieved land-use map shows kappa indices of 91, 79, and 76% agreement for location in comparison with the GMESUA data set at three levels of classification. Furthermore, the attributes of the two data sets match at 81, 67, and 65%. The results demonstrate that this approach opens a promising avenue to integrate freely available VGI to map land-use patterns for environmental planning purposes.

Mining urban land-use patterns from volunteered geographic information by means of genetic algorithms and artificial neural networks

In the context of OpenStreetMap (OSM), spatial data quality, in particular completeness, is an essential aspect of its fitness for use in specific applications, such as planning tasks. To mitigate the effect of completeness errors in OSM, this study proposes a methodological framework for predicting by means of OSM urban areas in Europe that are currently not mapped or only partially mapped. For this purpose, a machine learning approach consisting of artificial neural networks and genetic algorithms is applied. Under the premise of existing OSM data, the model estimates missing urban areas with an overall squared correlation coefficient (R 2) of 0.589. Interregional comparisons of European regions confirm spatial heterogeneity in the model performance, whereas the R 2 ranges from 0.129 up to 0.789. These results show that the delineation of urban areas by means of the presented methodology depends strongly on location.

Data-Driven Regionalization of Housing Markets

This article presents a data-driven framework for housing market segmentation. Local marginal house price surfaces are investigated by means of mixed geographically weighted regression and are reduced to a set of principal component maps, which in turn serve as input for spatial regionalization. The out-of-sample prediction error of a hedonic pricing model is applied to determine a “near-optimal” number of spatially coherent and homogeneous submarkets. The usefulness of this method is demonstrated with a detailed data set for the Austrian housing market. The results provide evidence that submarkets must always be considered, however they are defined, and that the proposed submarket taxonomy on a regional level significantly improves predictive quality compared to (1) a traditional pooled model, (2) a model that uses an ad hoc submarket definition based on administrative units, and (3) a model incorporating an alternative submarket definition on the basis of aspatial k-means clustering. Moreover, it is concluded that the Austrian housing market is characterized by regional determinants and that geography is the most important component determining the house prices.

Exploration of unstructured narrative crime reports: an unsupervised neural network and point pattern analysis approach

Crime intelligence analysis and criminal investigations are increasingly making use of geospatial methodologies to improve tactical and strategic decision-making. However, the full potential of geospatial technologies is yet to be exploited. In particular, geospatial technology currently applied by law enforcement is somewhat limited in handling the increasing volume of police recorded and relatively unstructured narrative crime reports, such as observations and interviews of eyewitnesses, the general public, or other relevant persons. The main objective of this research is to promote text mining, particularly the self-organizing map algorithm and its visualization capabilities, in combination with point pattern analysis, to explore the value of otherwise hidden information in a geographical context and to gain further insight into the complex behavior of the geography of crime. This methodological approach is applied to a high-profile and still unsolved homicide series in the city of Jennings, Louisiana. In a collaborative effort with the Jennings Police Task Force, the analysis is based upon a range of information sources, including email correspondence, transcribed face-to-face interviews, and phone calls that have been stored as “Information Packages” in the Orion database, which is maintained by the Federal Bureau of Investigation. Close to 200 individual information packages related to Necole Guillory, the eighth and last victim whose dead and dumped body was discovered in August 2009, are analyzed and resulted in new geographic patterns and relationships previously unknown to the Task Force.

Performance analysis of radial basis function networks and multi-layer perceptron networks in modeling urban change: a case study

The majority of cities are rapidly growing. This makes the monitoring and modeling of urban change’s spatial patterns critical to urban planners, decision makers, and environment protection activists. Although a wide range of methods exists for modeling and simulating urban growth, machine learning (ML) techniques have received less attention despite their potential for producing highly accurate predictions of future urban extents. The aim of this study is to investigate two ML techniques, namely radial basis function network (RBFN) and multi-layer perceptron (MLP) networks, for modeling urban change. By predicting urban change for 2010, the models’ performance is evaluated by comparing results with a reference map and by using a set of pertinent statistical measures, such as average spatial distance deviation and figure of merit. The application of these techniques employs the case study area of Mumbai, India. The results show that both models, which were tested using the same explanatory variables, produced promising results in terms of predicting the size and extent of future urban areas. Although a close match between RBFN and MLP is observed, RBFN demonstrates higher spatial accuracy of prediction. Accordingly, RBFN was utilized to simulate urban change for 2020 and 2030. Overall, the study provides evidence that RBFN is a robust and efficient ML technique and can therefore be recommended for land use change modeling.

Hierarchical self-organizing maps for clustering spatiotemporal data

Spatial sciences are confronted with increasing amounts of high-dimensional data. These data commonly exhibit spatial and temporal dimensions. To explore, extract, and generalize inherent patterns in large spatiotemporal data sets, clustering algorithms are indispensable. These clustering algorithms must account for the distinct special properties of space and time to outline meaningful clusters in such data sets. Therefore, this research develops a hierarchical method based on self-organizing maps. The hierarchical architecture permits independent modeling of spatial and temporal dependence. To exemplify the utility of the method, this research uses an artificial data set and a socio-economic data set of the Ostregion, Austria, from the years 1961 to 2001. The results for the artificial data set demonstrate that the proposed method produces meaningful clusters that cannot be achieved when disregarding differences in spatial and temporal dependence. The results for the socio-economic data set show that the proposed method is an effective and powerful tool for analyzing spatiotemporal patterns in a regional context.

Contextual neural gas for spatial clustering and analysis

This study aims to introduce contextual Neural Gas (CNG), a variant of the Neural Gas algorithm, which explicitly accounts for spatial dependencies within spatial data. The main idea of the CNG is to map spatially close observations to neurons, which are close with respect to their rank distance. Thus, spatial dependency is incorporated independently from the attribute values of the data. To discuss and compare the performance of the CNG and GeoSOM, this study draws from a series of experiments, which are based on two artificial and one real-world dataset. The experimental results of the artificial datasets show that the CNG produces more homogenous clusters, a better ratio of positional accuracy, and a lower quantization error than the GeoSOM. The results of the real-world dataset illustrate that the resulting patterns of the CNG are theoretically more sound and coherent than that of the GeoSOM, which emphasizes its applicability for geographic analysis tasks.

Clustering Contextual Neural Gas: A New Approach for Spatial Planning and Analysis Tasks

Spatial clustering is a method that can reveal structures and identify groupings in large spatial data sets, which is in particular useful for spatial planning and analysis tasks. A recent and powerful clustering algorithm for spatial data is contextual neural gas (CNG). The CNG algorithm is closely related to the basic self-organizing map algorithm but additionally takes spatial dependence into account. However, like most clustering algorithms, CNG requires the analyst to specify the number of clusters beforehand. Even though the chosen number of clusters critically affects the results of the clustering, it is unclear how to determine it. This study introduces a new method which combines CNG, the learning of the CNG’s topology, and graph clustering. It can be used to cluster spatial data without any prior knowledge of present clusters in the data. The proposed method is in particular useful for spatial planning and analysis tasks, because it provides means to find groupings in the data and identify homogeneous regions. To evaluate the method, this study draws from two experiments which are based on a synthetic and a real-world data set. The results of the synthetic data set show that it can correctly identify clusters in a predefined setting. The results of the real-world data set demonstrate that the proposed method outlines meaningful and theoretically sound regions.

Big data in Geohazard; pattern mining and large scale analysis of landslides in Iran

In this paper, we clustered and analyzed landslides and investigated their underlying driving forces at two levels, country and cluster, all over Iran. Considering 12 conditioning factors, the landslides were clustered into nine relatively homogeneous regions using the Contextual Neural Gas (CNG) algorithm. Next, their underlying driving forces were ranked using the Random Forest (RF) algorithm at country and cluster levels. Our results indicate that the mechanisms for landslide occurrence varied for each cluster and that driving forces of the landslides operated differently at a country level compared to the cluster level. Moreover, slope, altitude, average annual rainfall, and distance to the main roads were identified as the most important causes of landslides within all clusters. Thus, for effective management and modelling landslides on a large scale, the variation in the functionality of effective factors should be considered.