Mapping forest alliances and associations using fuzzy systems and nearest neighbor classifiers

The study and management of biological communities depends on systems of classification and mapping for the organization and communication of resource information. Recent advances in remote sensing technology may enable the mapping of forest plant associations using image classification techniques. But few areas outside Europe have alliances and associations described in detail sufficient to support remote sensing-based modeling. Northwestern Montana has one of the few completed plant association classifications in the United States compliant with the recently established National Vegetation Classification system. This project examined the feasibility of mapping forest plant associations using Landsat Enhanced Thematic Mapper Plus data and advanced remote sensing technology and image classification techniques.Suitable reference data were selected from an extensive regional database of plot records. Fifteen percent of the plot samples were reserved for validation of map products, the remainder of plots designated as training data for map modeling. Key differentiae for image classification were identified from a suite of spectral and biophysical variables. Fuzzy rules were formulated for partitioning physiognomic classes in the upper levels of our image classification hierarchy. Nearest neighbor classifiers were developed for classification of lower levels (alliances and associations), where spectral and biophysical contrasts are less distinct.Maps were produced to reflect nine forest alliances and 24 associations across the study area. Error matrices were constructed for each map based on stratified random selections of map validation samples. Accuracy for the alliance map was estimated at 60%. Association classifiers provide between 54 and 86% accuracy within their respective alliances. Alternative techniques are proposed for aggregating classes and enhancing decision tree classifiers to model alliances and associations for interior forest types.     

Combining classifiers using nearest decision prototypes

We present a new classifier fusion method to combine soft-level classifiers with a new approach, which can be considered as a generalized decision templates method. Previous combining methods based on decision templates employ a single prototype for each class, but this global point of view mostly fails to properly represent the decision space. This drawback extremely affects the classification rate in such cases: insufficient number of training samples, island-shaped decision space distribution, and classes with highly overlapped decision spaces. To better represent the decision space, we utilize a prototype selection method to obtain a set of local decision prototypes for each class. Afterward, to determine the class of a test pattern, its decision profile is computed and then compared to all decision prototypes. In other words, for each class, the larger the numbers of decision prototypes near to the decision profile of a given pattern, the higher the chance for that class. The efficiency of our proposed method is evaluated over some well-known classification datasets suggesting superiority of our method in comparison with other proposed techniques.     

Reweighting anthropometric data using a nearest neighbour approach

When designing products and environments, detailed data on body size and shape are seldom available for the specific user population. One way to mitigate this issue is to reweight available data such that they provide an accurate estimate of the target population of interest. This is done by assigning a statistical weight to each individual in the reference data, increasing or decreasing their influence on statistical models of the whole. This paper presents a new approach to reweighting these data. Instead of stratified sampling, the proposed method uses a clustering algorithm to identify relationships between the detailed and reference populations using their height, mass, and body mass index (BMI). The newly weighted data are shown to provide more accurate estimates than traditional approaches. The improved accuracy that accompanies this method provides designers with an alternative to data synthesis techniques as they seek appropriate data to guide their design practice.Practitioner Summary: Design practice is best guided by data on body size and shape that accurately represents the target user population. This research presents an alternative to data synthesis (e.g. regression or proportionality constants) for adapting data from one population for use in modelling another.     

Design of supervised classifiers using Boolean neural networks

In this paper we present two supervised pattern classifiers designed using Boolean neural networks. They are: 1) nearest-to-an-exemplar classifier; and 2) Boolean k-nearest neighbor classifier. The emphasis during the design of these classifiers was on simplicity, robustness, and the ease of hardware implementation. The classifiers use the idea of radius of attraction to achieve their goal. Mathematical analysis of the algorithms presented in the paper is done to prove their feasibility. Both classifiers are tested with well-known binary and continuous feature valued data sets yielding results comparable with those obtained by similar existing classifiers     

On visualization and aggregation of nearest neighbor classifiers

Nearest neighbor classification is one of the simplest and most popular methods for statistical pattern recognition. A major issue in k-nearest neighbor classification is how to find an optimal value of the neighborhood parameter k. In practice, this value is generally estimated by the method of cross-validation. However, the ideal value of k in a classification problem not only depends on the entire data set, but also on the specific observation to be classified. Instead of using any single value of k, this paper studies results for a finite sequence of classifiers indexed by k. Along with the usual posterior probability estimates, a new measure, called the Bayesian measure of strength, is proposed and investigated in this paper as a measure of evidence for different classes. The results of these classifiers and their corresponding estimated misclassification probabilities are visually displayed using shaded strips. These plots provide an effective visualization of the evidence in favor of different classes when a given data point is to be classified. We also propose a simple weighted averaging technique that aggregates the results of different nearest neighbor classifiers to arrive at the final decision. Based on the analysis of several benchmark data sets, the proposed method is found to be better than using a single value of k.     

Optimized skyline queries on road networks using nearest neighbors

Skyline queries are used with data extensive applications, such as mobile location-based services, to support multi-criteria decision-making and to prune the data space by returning the most “interesting” data points. Most interesting data points are the points, which are not dominated by any other point. Spatial network skyline query is a subset of the skyline query problem where data points are nodes in a road network and the attributes of the data points are network distance relative to a set of query points. Spatial network skyline query’s problem is the need to calculate the attributes with an expensive distance calculation operation. Previous works (Deng et al. Proceedings of the 23th international conference on data engineering, 796–805, 2007), Sharifzadeh et al. Proceedings of the 32nd international conference on very large databases, 751–762, 2009) that addressed this problem involved extensive network distance calculation between the query points and data points. A new algorithm that requires a remarkably less number of network distance calculations is proposed in this work. Our approach uses a progressive nearest neighbor algorithm to minimize the set of candidates then evaluates those candidates by only comparing them to a subset of discovered skyline points. Experiments showed the effectiveness of our algorithm compared to previous works.     

Comparison of the nearest feature classifiers for face recognition

This paper presents an experimental comparison of the nearest feature classifiers, using an approach based on binomial tests in order to evaluate their strengths and weaknesses. In addition, classification accuracies and the accuracy-dimensionality tradeoff have been considered as comparison criteria. We extend two of the nearest feature classifiers to label the query point by a majority vote of the samples. Comparisons were carried out for face recognition using ORL database. We apply the eigenface representation for feature extraction. Experimental results showed that even though the classification accuracy of k-NFP outperforms k-NFL in some dimensions, these rate differences do not have statistical significance.     

Mapping tropical forest structure in southeastern Madagascar using remote sensing and artificial neural networks

Tropical forest condition has important implications for biodiversity, climate change and human needs. Structural features of forests can serve as useful indicators of forest condition and have the potential to be assessed with remotely sensed imagery, which can provide quantitative information on forest ecosystems at high temporal and spatial resolutions. Herein, we investigate the utility of remote sensing for assessing, predicting and mapping two important forest structural features, stem density and basal area, in tropical, littoral forests in southeastern Madagascar. We analysed the relationships of basal area and stem density measurements to the Normalised Difference Vegetation Index (NDVI) and radiance measurements in bands 3, 4, 5 and 7 from the Landsat Enhanced Thematic Mapper Plus (ETM+). Strong relationships were identified among all of the individual bands and field based measurements of basal area (p<0.01) while there were weak and insignificant relationships among spectral response and stem density measurements. NDVI was not significantly correlated with basal area but was strongly and significantly correlated with stem density (r=−0.69, p<0.01) when using a subset of the data, which represented extreme values. We used an artificial neural network (ANN) to predict basal area from radiance values in bands 3, 4, 5 and 7 and to produce a predictive map of basal area for the entire forest landscape. The ANNs produced strong and significant relationships between predicted and actual measures of basal area using a jackknife method (r=0.79, p<0.01) and when using a larger data set (r=0.82, p<0.01). The map of predicted basal area produced by the ANN was assessed in relation to a pre-existing map of forest condition derived from a semi-quantitative field assessment. The predictive map of basal area provided finer detail on stand structural heterogeneity, captured known climatic influences on forest structure and displayed trends of basal area associated with degree of human accessibility. These findings demonstrate the utility of ANNs for integrating satellite data from the Landsat ETM+ spectral bands 3, 4, 5 and 7 with limited field survey data to assess patterns in basal area at the landscape scale.     

Delineation of forest/nonforest land use classes using nearest neighbor methods

The k-Nearest Neighbor (kNN) method of forest attribute estimation and mapping has become an integral part of national forest inventory methods in Finland in the last decade. This success of kNN method in facilitating multi-source inventory has encouraged trials of the method in the Great Lakes Region of the United States. Here we present results from applying the method to Landsat TM and ETM+ data and land cover data collected by the USDA Forest Service's Forest Inventory and Analysis (FIA) program. In 1999, the FIA program in the state of Minnesota moved to a new annual inventory design to reach its targeted full sampling intensity over a 5-year period. This inventory design also utilizes a new 4-subplot cluster plot configuration. Using this new plot design together with 1 year of field plot observations, the kNN classification of forest/nonforest/water achieved overall accuracies ranging from 87% to 91%. Our analysis revealed several important behavioral features associated with kNN classification using the new FIA sample plot design. Results demonstrate the simplicity and utility of using kNN to produce FIA defined forest/nonforest/water classifications.     

Assessing the performance of a modified S-transform with probabilistic neural network,support vector machine and nearest neighbour classifiers for single and multiple power quality disturbances identification

Neural Computing and Applications - This paper presents a substantial assessment between a modified S-transform (MST) and the original S-transform (OST) for the identification of single and...     

Training sets and a priori probabilities with the nearest neighbour method of pattern recognition

This paper considers whether the nearest neighbour (NN) classifier takes proper account of a priori class probabilities. If the frequencies with which the different classes arise naturally in the whole population of patterns are retained during training, it is found that a priori probabilities are automatically incorporated: this will not be so if (for example) equal numbers of training patterns are selected from each class.     

An analysis of how training data complexity affects the nearest neighbor classifiers

The k-nearest neighbors (k-NN) classifier is one of the most popular supervised classification methods. It is very simple, intuitive and accurate in a great variety of real-world domains. Nonetheless, despite its simplicity and effectiveness, practical use of this rule has been historically limited due to its high storage requirements and the computational costs involved. On the other hand, the performance of this classifier appears strongly sensitive to training data complexity. In this context, by means of several problem difficulty measures, we try to characterize the behavior of the k-NN rule when working under certain situations. More specifically, the present analysis focuses on the use of some data complexity measures to describe class overlapping, feature space dimensionality and class density, and discover their relation with the practical accuracy of this classifier.

---

Defying the gravity of learning curve: a characteristic of nearest neighbour anomaly detectors

Conventional wisdom in machine learning says that all algorithms are expected to follow the trajectory of a learning curve which is often colloquially referred to as ‘more data the better’. We call this ‘the gravity of learning curve’, and it is assumed that no learning algorithms are ‘gravity-defiant’. Contrary to the conventional wisdom, this paper provides the theoretical analysis and the empirical evidence that nearest neighbour anomaly detectors are gravity-defiant algorithms.     

Classification of boreal forest cover types using SAR images

Mapping forest cover types in the boreal ecosystem is important for understanding the processes governing the interaction of the surface with the atmosphere. In this paper, we report the results of the land-cover classification of the SAR (synthetic aperture radar) data acquired during the Boreal Ecosystem Atmospheric Study's intensive field campaigns over the southern study area near Prince Albert, Canada. A Bayesian maximum a posteriori classifier was applied on the national Aeronautics and Space Administration/Jet Propulsion Laboratory airborne SAR images covering the region during the peak of the growing season in July 1994. The approach is supervised in the sense that a combination of field data and existing land-cover maps are used to develop training areas for the desired classes. The images acquired were first radiometrically and absolutely calibrated, the incidence angle effect in airborne images was corrected to an acceptable accuracy, and the images were used in a mosaic form and geocoded and georeferenced with an existing land-cover map for validation purposes. The results show that SAR images can be classified into dominant forest types such as jack pine, black spruce, trembling aspen, clearing, open water, and three categories of mixed strands with better than 90% accuracy. The unispecies stands such as jack pine and black spruce are separated with 98% accuracy, but the accuracy of mixed coniferous and deciduous stands suffers from confusing factors such as varying species composition, surface moisture, and understory effects. To satisfy the requirements of process models, the number of cover types was reduced from eight to five general classes of conifer wet, conifer dry, mixed deciduous, disturbed, and open water. Reduction of classes improved the overall accuracy of the classification over the entire region from 77% to 92%.     

Statistical computation of feature weighting schemes through data estimation for nearest neighbor classifiers

The Nearest Neighbor rule is one of the most successful classifiers in machine learning. However, it is very sensitive to noisy, redundant and irrelevant features, which may cause its performance to deteriorate. Feature weighting methods try to overcome this problem by incorporating weights into the similarity function to increase or reduce the importance of each feature, according to how they behave in the classification task. This paper proposes a new feature weighting classifier, in which the computation of the weights is based on a novel idea combining imputation methods – used to estimate a new distribution of values for each feature based on the rest of the data – and the Kolmogorov–Smirnov nonparametric statistical test to measure the changes between the original and imputed distribution of values. This proposal is compared with classic and recent feature weighting methods. The experimental results show that our feature weighting scheme is very resilient to the choice of imputation method and is an effective way of improving the performance of the Nearest Neighbor classifier, outperforming the rest of the classifiers considered in the comparisons.     

Evaluation of most similar neighbour and random forest methods for imputing forest inventory variables using data from target and auxiliary stands

We report the results from modelling standing volume, above-ground biomass and stem count with the aim of exploring the potential of two non-parametric approaches to estimate forest attributes. The models were built based on spectral and 3D information extracted from airborne optical and laser scanner data. The survey was completed across two geographically adjacent temperate forest sites in southwestern Germany, using spatially and temporally comparable remote-sensing data collected by similar instruments. Samples from the auxiliary reference stands (called off-site samples) were combined with random, random stratified and systematically stratified samples from the target area for prediction of standing volume, above-ground biomass and stem count in the target area. A range of combinations was used for the modelling process, comprising the most similar neighbour (MSN) and random forest (RF) imputation methods, three sampling designs and two predictor subset sizes. An evolutionary genetic algorithm (GA) was applied to prune the predictor variables. Diagnostic tools, including root mean square error (RMSE), bias and standard error of imputation, were employed to evaluate the results. The results showed that RF produced more accurate results than MSN (average improvement of 3.5% for a single-neighbour case with selected predictors), yet was more biased than MSN (average bias of 5.13% with RF compared to 2.44% with MSN for stem volume in a single-neighbour case with selected predictors). Combining systematically stratified auxiliary samples from the target data set with the reference data set yielded more accurate results compared to those from random and stratified random samples. Combining additional data was most influential when an intensity of up to 40% of supplementary samples was appended to the reference set. The use of GA-selected predictors resulted in reduced bias of the models. By means of bootstrap simulations of RMSE, the simulations were shown to lie within the applied non-parametric confidence intervals. The achieved results are concluded to be helpful for modelling the mentioned forest attributes by means of airborne remote-sensing data.     

Efficient exact k-flexible aggregate nearest neighbor search in road networks using the M-tree

The Journal of Supercomputing - This study proposes an efficient exact k-flexible aggregate nearest neighbor (k-FANN) search algorithm in road networks using the M-tree. The state-of-the-art...     

Inversion of a forest backscatter model using neural networks

A neural network approach was used to develop acccurate algorithms for inverting a complex forest backscatter model. The model combines a forest growth model with a radar backscatter model. The forest growth model captures natural variations of forest stands (e.g., growth, regeneration, death, multiple species and competition for light). This model was used to produce vegetation structure data typical of transitional/northern boreal hardwood forests in Maine. These data supplied inputs to the radar backscatter model which simulated the polarimetric radar backscatter (C, L, P, X bands) above the forests. Using these simulated data, various neural networks were trained with inputs of different backscatter bands and output parameters of above ground biomass, total number of trees, mean tree height and mean tree age. These trained neural networks act as efficient algorithms for inverting the complex forest backscatter model. The accuracies (r.m.s. and R2 values) for inferring various parameters from radar backscatter were above ground biomass (1.6kg m -2, 0.94), number of trees (48 ha -1, 0.94), tree height (0.47 m, 0.88) and tree age (24.0 years, 0.83). The networks that used only AIRSAR bands (C, L, P) had a high degree of accuracy. The inclusion of the X band with the AIRSAR bands did not seem to increase significantly the accuracy of the networks. The networks that used only the C and L bands still had a relatively high degree of accuracy for all forest parameter (R2 values from 0.75 to 0.91). Modest accuracies (R2 values from 0.65 to 0.84) were obtained with networks that used only the L band and poor accuracies (R2 values from 0.36 to 0.46) were obtained with networks that used only the C band. Several networks were shown to be relatively insensitive to the addition of random noise to radar backscatter. The results demonstrate that complex, forest backscatter models can be efficiently inverted using neural networks that use only radar backscatter data.     

Augmentation of a nearest neighbour clustering algorithm with a partial supervision strategy for biomedical data classification

Abstract: In this paper, a partial supervision strategy for a recently developed clustering algorithm, the nearest neighbour clustering algorithm (NNCA), is proposed. The proposed method (NNCA-PS) offers classification capability with a smaller amount of a priori knowledge, where a small number of data objects from the entire data set are used as labelled objects to guide the clustering process towards a better search space. Experimental results show that NNCA-PS gives promising results of 89% sensitivity at 95% specificity when used to segment retinal blood vessels, and a maximum classification accuracy of 99.5% with 97.2% average accuracy when applied to a breast cancer data set. Comparisons with other methods indicate the robustness of the proposed method in classification. Additionally, experiments on parallel environments indicate the suitability and scalability of NNCA-PS in handling larger data sets.