Average Misclassification Error Research Articles

An automatic purse-seine set type classification algorithm to inform tropical tuna management

Management of large-scale tuna purse-seine fisheries often involves consideration of set type because the different set types yield different target species and bycatch species compositions, and may therefore impact those populations differently. Using the ensemble method random forests for classification, annual set type classification algorithms were developed based on a suite of operational characteristics, and catch and bycatch information, collected by onboard observers for the tropical tuna purse-seine fishery in the eastern Pacific Ocean (EPO) during 2010–2019. Two types of algorithms were built for each year: 1) a 2-class algorithm for sets on unassociated tuna schools (NOA) and sets on floating-object-associated tuna (OBJ); and, 2) a 3-class algorithm for all three types of sets that occur in the EPO, which includes sets on marine mammal-associated tuna (DEL). Results indicated that such data can be used to reliably distinguish between purse-seine sets types, highlighting inherent operational and ecological differences among them in the EPO. For the 2-class algorithms, the average misclassification error rates were 3.3% for NOA sets and 4.5% for OBJ sets. For the 3-class algorithms, the average misclassification error rates were 2% for DEL sets, 10.6% for NOA sets and 4.6% for OBJ sets. Comparison of results for NOA sets from the 2-class and 3-class algorithms suggests that the higher NOA-set misclassification error rate for the 3-class algorithm is due to a difficulty in distinguishing NOA sets from DEL sets with the predictors used in this analysis. The most useful bycatch information for predicting set type included amounts of dorado, wahoo, small fish species, such as triggerfishes, and silky sharks, illustrating the importance of bycatch data collection by observers, even for species of no commercial value. In the case of the 2-class algorithms, the annual misclassification error rates for NOA and OBJ sets were not only small, but also fairly stable over the 10-year period, even in the presence of the strong 2015 – 2016 El Niño event. This suggests that the methodology is robust and could be used to validate set type determinations based on other criteria, such as distance to a floating object (or a fish aggregating device), and to verify data quality, more generally.

Fast Single-Parameter Energy Function Thresholding for Image Segmentation Based on Region Information

To solve the problems of image threshold segmentation based on weak continuous constraint theory, the running time is long, and the two parameters need to be selected manually, and therefore a fast single-parameter energy function thresholding for image segmentation based on region information (FSEFTISRI) is proposed in this paper. The proposed FSEFTISRI algorithm uses simple linear iterative clustering (SLIC) technology to pre-block the image, extract the image super-pixels, and then map the image super-pixels to the interval type-2 fuzzy set (IT2FS), so as to construct the single-parameter energy function to search the optimal threshold, and adaptively select the penalty parameters in the energy function through the class uncertainty theory. On a non-destructive testing (NDT) database and Berkeley segmentation datasets and benchmarks (BSDS), the proposed FSEFTISRI is compared with five related algorithms. The average misclassification error (ME) of the proposed FSEFTISRI algorithm on NDT and BSDS are 0.0466 and 0.0039, respectively. The results show that the proposed FSEFTISRI has acquired more satisfactory results in visual effect and evaluation index, and the running time of the proposed FSEFTISRI algorithm is shorter, which shows the effectiveness of the proposed FSEFTISRI.

Computer-Aided System of the Mandibular Cortical Bone Porosity Assessment on Digital Panoramic Radiographs.

The loss of bone mineral density (BMD) in various sites of the body, including the mandible, is the main sign of osteoporosis. Thus, the computer-aided diagnosis (CAD) system was developed for bone density assessment and patients were classified into normal, osteopenia, and osteoporosis groups using a digital panoramic radiograph. Data of dental panoramic radiographs and corresponding BMD assessments from 123 postmenopausal women were collected. For the proposed CAD system test, regions of interest (ROI) that were located below the left and right mental foramen on dental panoramic radiographs were determined. The width and texture of the mandibular cortical bone in each ROI were used to classify the data into normal, osteopenia, and osteoporosis classes. The width of the mandibular cortical was measured using the polynomial fitting method. The texture feature of the cortical bone is obtained by calculating the average value of the grayscale intensity of cortical bone. The classification result was obtained by using a multiclass support vector machine. The experimental results using 10-fold cross-validation showed that the proposed system achieved an average accuracy of 86.50% for osteoporosis classification on dental panoramic radiographs. The average misclassification error and relative foreground area error of the segmentation process were 5.21 and 12.98%, respectively. From the analysis of the cortical width measurement process, highest average mandibular cortical width (MCW) was found in the normal patient category compared with the other classes. This research showed that the proposed computer-aided system can be used for osteoporosis and osteopenia assessment by measuring the MCW and texture on dental panoramic radiographs with the average system accuracy of 89.52%.

Double Thresholding with Sine Entropy for Thermal Image Segmentation

Traditional thresholding methods are often used for image segmentation of real images. However, due to distinct characteristics of infrared thermal images, it is difficult to ensure an optimal image segmentation using the traditional thresholding algorithms, and therefore, sometimes this can lead to over-segmentation, missing object information, and/or spurious responses in the output. To overcome these issues, we propose a new thresholding technique that makes use of the sine entropy-based criterion. Moreover, we build a double thresholding technique that makes use of two thresholds to get the final image thresholding result. Besides, we introduce the sine entropy concept as a supplement of the Shannon entropy in creating threshold-dependent criterion derived from the grayscale histogram. We found that the sine entropy is more robust in interpreting the strength of the long-range correlation in the gray levels compared to the Shannon entropy. We have experimented with our method on several infrared thermal images collected from standard image databases to describe the performance. On comparing with the state-of-art methods, the qualitative results from the experiments show that the proposed method achieves the best performance with an average sensitivity of 0.98 and an average misclassification error of 0.01, and second-best performance with an average sensitivity of 0.99 and an average specificity of 0.93.

Validation of the Cognitive Telephone Screening Instruments COGTEL and COGTEL+ in Identifying Clinically Diagnosed Neurocognitive Disorder Due to Alzheimer's Disease in a Naturalistic Clinical Setting.

Background: Telephone-based neurocognitive instruments embody valuable tools in identifying cognitive impairment in research settings and lately also in clinical contexts due to the pandemic crisis. The accuracy of the Cognitive Telephone Screening Instrument (COGTEL) in detecting mild- (MiND) and major (MaND) neurocognitive disorder has not been studied yet.Objective:Comparison of the utility of COGTEL and COGTEL+, which is enriched with orientation items, with the modified Mini-Mental State Examination (3MS) in detecting MiND and MaND due to Alzheimer’s disease (AD) and assessment of the impact of COGTEL face-to-face-versus telephone administration on individual performance.Methods:The study included 197 cognitively intact individuals (CI), being at least 45 years old, 95 and 65 patients with MiND and MaND due to AD, respectively. In 20 individuals COGTEL was administered both in face-to-face and telephone sessions. Statistical analyses included proportional odds logistic regression models, stratified repeated random subsampling used to recursive partitioning to training and validation set (70/30 ratio), and an appropriate F-test.Results:All studied instruments were significant predictors of diagnostic outcome, but COGTEL+ and 3MS explained more variance relative to the original COGTEL. Except for the validation regression models including COGTEL in which the average misclassification error slightly exceeded 15%, in all other cases the average misclassification errors (%) were lower than 15%. COGTEL administration modality was not related to systematic over- or underestimation of performance on COGTEL.Conclusion:COGTEL+ is a valuable instrument in detecting MiND and MaND and can be administered in face-to-face or telephone sessions.

Unsupervised classification of SDSS galaxy spectra

Context. Defining templates of galaxy spectra is useful to quickly characterise new observations and organise databases from surveys. These templates are usually built from a pre-defined classification based on other criteria. Aims. We present an unsupervised classification of 702 248 spectra of galaxies and quasars with redshifts smaller than 0.25 that were retrieved from the Sloan Digital Sky Survey (SDSS) database, release 7. Methods. The spectra were first corrected for redshift, then wavelet-filtered to reduce the noise, and finally binned to obtain about 1437 wavelengths per spectrum. The unsupervised clustering algorithm Fisher-EM, relying on a discriminative latent mixture model, was applied on these corrected spectra. The full set and several subsets of 100 000 and 300 000 spectra were analysed. Results. The optimum number of classes given by a penalised likelihood criterion is 86 classes, of which the 37 most populated gather 99% of the sample. These classes are established from a subset of 302 214 spectra. Using several cross-validation techniques we find that this classification agrees with the results obtained on the other subsets with an average misclassification error of about 15%. The large number of very small classes tends to increase this error rate. In this paper, we do an initial quick comparison of our classes with literature templates. Conclusions. This is the first time that an automatic, objective and robust unsupervised classification is established on such a large number of galaxy spectra. The mean spectra of the classes can be used as templates for a large majority of galaxies in our Universe.

Leaf Segmentation and Classification with a Complicated Background Using Deep Learning

The segmentation and classification of leaves in plant images are a great challenge, especially when several leaves are overlapping in images with a complicated background. In this paper, the segmentation and classification of leaf images with a complicated background using deep learning are studied. First, more than 2500 leaf images with a complicated background are collected and artificially labeled with target pixels and background pixels. Two-thousand of them are fed into a Mask Region-based Convolutional Neural Network (Mask R-CNN) to train a model for leaf segmentation. Then, a training set that contains more than 1500 training images of 15 species is fed into a very deep convolutional network with 16 layers (VGG16) to train a model for leaf classification. The best hyperparameters for these methods are found by comparing a variety of parameter combinations. The results show that the average Misclassification Error (ME) of 80 test images using Mask R-CNN is 1.15%. The average accuracy value for the leaf classification of 150 test images using VGG16 is up to 91.5%. This indicates that these methods can be used to segment and classify the leaf image with a complicated background effectively. It could provide a reference for the phenotype analysis and automatic classification of plants.

Comparison of Error of Classification Associated With Gamma Distribution Using Some Selected Discriminant Functions

This study examined the error of classification associated with the gamma distribution for the linear discriminant analysis, the logistic discriminant analysis and the quadratic discriminant analysis. Data were simulated data from Gamma distribution for sample size 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100. Fifty simulation for each of the sample size was performed for the methods considered in this study. The findings of the study revealed that for the gamma distribution, the logistic discriminant analysis method performed better followed by the linear discriminant analysis and then the quadratic discriminant analysis. This is because the logistic discriminant analysis method has the least average misclassification error rate across the sample sizes followed by the linear discriminant analysis method while the quadratic discriminant method performed least for gamma distribution.

Imbalanced TSK Fuzzy Classifier by Cross-Class Bayesian Fuzzy Clustering and Imbalance Learning

In this paper, a novel construction algorithm called imbalanced Takagi–Sugeno–Kang fuzzy classifier (IB-TSK-FC) for the TSK fuzzy classifier is presented to improve the classification performance and rule-based interpretability for imbalanced datasets. IB-TSK-FC consists of two components: 1) a cross-class Bayesian fuzzy clustering algorithm (BF3C) and 2) an imbalance learning algorithm. In order to achieve high interpretability, BF3C is developed to determine an appropriate number of fuzzy rules and identify antecedent parameters of fuzzy rules from the perspective of the probabilistic model. In addition to inheriting the distinctive advantage of Bayesian fuzzy clustering that the number of clusters can be estimated in the framework of Bayesian inference, BF3C considers repulsion forces between cluster centers belonging to different classes, and uses an alternating iterative strategy to obtain more interpretable antecedent parameters for imbalanced datasets. In order to improve the classification performance for imbalanced datasets, an imbalance learning algorithm is derived to estimate consequent parameters of fuzzy rules on the basis of the weighted average misclassification error. Comprehensive experiments on synthetic and UCI datasets demonstrate the effectiveness of the proposed IB-TSK-FC algorithm.

MULTIVARIATE STATISTICAL ANALYSIS OF COLEOPTERA SPECTRAL REFLECTANCE

The insect order Coleoptera, commonly known as beetles, comprises 40% of all insects which in turn account for half of all identified animal species alive today. Coleopterans frequently have large elytra (the hardened front wings) that can have a wide range of colors. Spectral reflectance readings from these elytra may be used to uniquely identify coleopteran taxonomic groups. Multiple samples of eleven species of wood boring beetles were selected from the University of Idaho William Barr Entomology Museum. Spectrometer readings for each specimen were then fit to normal distribution mixture models to identify multiple peak reflectance wavelengths. Eighteen prominent peaks were identified across all taxonomic groups and genders creating a multivariate response structure. Multivariate statistical procedures including principal component and discriminant analyses were employed to assess the differentiation of taxonomic groups and genders based on spectral reflectance. The first three axes of the principal component analysis accounted for 96% of the variation and provided a clear clustering of genus and gender for a subset of taxonomic groups. The linear discriminant analysis under an assumption of multivariate normality provided a distinct classification of taxonomic groups resulting in an overall 4% misclassification rate; while the nearest neighbor discriminant analysis with a proportional prior gave an overall error rate of 5.2%. Internal bootstrap validation of the latter discriminant model yielded an average error rate of 3.5%. An external cross validation of the same model, conducted on independent samples of the same species with new individuals resulted in an average misclassification error rate of only 6.5%. Given the low error rates of misclassification, such multivariate statistical approaches are recommended for analysis of spectral reflectance in Coleoptera and other similar insect groups.

Assessing stream bank condition using airborne LiDAR and high spatial resolution image data in temperate semirural areas in Victoria, Australia

Stream bank condition is an important physical form indicator for streams related to the environmental condition of riparian corridors. This research developed and applied an approach for mapping bank condition from airborne light detection and ranging (LiDAR) and high-spatial resolution optical image data in a temperate forest/woodland/urban environment. Field observations of bank condition were related to LiDAR and optical image-derived variables, including bank slope, plant projective cover, bank-full width, valley confinement, bank height, bank top crenulation, and ground vegetation cover. Image-based variables, showing correlation with the field measurements of stream bank condition, were used as input to a cumulative logistic regression model to estimate and map bank condition. The highest correlation was achieved between field-assessed bank condition and image-derived average bank slope (R2 1/4 0.60, n 1/4 41), ground vegetation cover (R2 1/4 0.43, n 1/4 41), bank width/height ratio (R2 1/4 0.41, n 1/4 41), and valley confinement (producer's accuracy 1/4 100%, n 1/4 9). Crossvalidation showed an average misclassification error of 0.95 from an ordinal scale from 0 to 4 using the developed model. This approach was developed to support the remotely sensed mapping of stream bank condition for 26,000 km of streams in Victoria, Australia, from 2010 to 2012.

Development of a hybrid support vector machine and genetic algorithm model for regime identification of slurry transport in pipelines

AbstractFour distinct regimes (namely sliding bed, saltation, heterogeneous suspension, and homogeneous suspension) were found existent in slurry flow in a pipeline depending upon the average velocity of flow. In the literature, a few numbers of correlations have been proposed for identification of these regimes in slurry pipelines.Regime identification is important for slurry pipeline design as it is the prerequisite to apply different pressure drop correlations in different regimes. However, available correlations fail to predict the regime over a wide range of conditions. Based on a databank of around 800 measurements collected from the open literature, a method has been proposed to identify the regime using support vector machine (SVM) modeling. The method incorporates hybrid support vector machine and genetic algorithm technique (SVM‐GA) for efficient tuning of SVM meta‐parameters. Statistical analysis showed that the proposed method has an average misclassification error (AARE) of 0.03%. A comparison with selected correlations in the literature showed that the developed SVM‐GA method noticeably improved prediction of regimes over a wide range of operating conditions, physical properties, and pipe diameters. Copyright © 2009 Curtin University of Technology and John Wiley & Sons, Ltd.

Regime identification of slurry transport in pipelines: A novel modelling approach using ANN & differential evolution

Four distinct regimes were found existent (namely sliding bed, saltation, heterogeneous suspension and homogeneous suspension) in slurry flow in pipeline depending upon the average velocity of flow. In the literature, few numbers of correlations has been proposed for identification of these regimes in slurry pipelines. Regime identification is important for slurry pipeline design as they are the prerequisite to apply different pressure drop correlation in different regime. However, available correlations fail to predict the regime over a wide range of conditions. Based on a databank of around 800 measurements collected from the open literature, a method has been proposed to identify the regime using artificial neural network (ANN) modeling. The method incorporates hybrid artificial neural network and differential evolution technique (ANN - DE) for efficient tuning of ANN Meta parameters. Statistical analysis showed that the proposed method has an average misclassification error of 0.03%. A comparison with selected correlations in the literature showed that the developed ANN - DE method noticeably improved prediction of regime over a wide range of operating conditions, physical properties, and pipe diameters. .

Development of a Hybrid Artificial Neural Network and Genetic Algorithm Model for Regime Identification of Slurry Transport in Pipelines

Four distinct regimes were found existent (namely sliding bed, saltation, heterogeneous suspension and homogeneous suspension) in slurry flow in pipeline depending upon the average velocity of flow. In the literature, few numbers of correlations has been proposed for identification of these regimes in slurry pipelines. Regime identification is important for slurry pipeline design as they are the prerequisite to apply different pressure drop correlation in different regime. However, available correlations fail to predict the regime over a wide range of conditions. Based on a databank of around 800 measurements collected from the open literature, a method has been proposed to identify the regime using artificial neural network (ANN) modeling. The method incorporates hybrid artificial neural network and genetic algorithm technique (ANN-GA) for efficient tuning of ANN meta parameters. Statistical analysis showed that the proposed method has an average misclassification error of 0.03%. A comparison with selected correlations in the literature showed that the developed ANN-GA method noticeably improved prediction of regime over a wide range of operating conditions, physical properties, and pipe diameters.

Comparison of misclassification rates of search partition analysis and other classification methods

Search partition analysis (SPAN) is a method to develop classification rules based on Boolean expressions. The performance of SPAN is compared against the trials reported by Lim et al. of 33 other methods of classification, including tree, neural network and regression methods on 16 data sets, most of which were health related. Each data set was augmented with noise variables in further trials. Lim et al. assessed the performance of the methods by estimates of misclassification rate, either cross-validated or test sample based. In this paper, the same data sets are analysed by SPAN and misclassification rates of the SPAN classifiers are estimated. Comparison is made of the performance of SPAN against the other methods that were considered by Lim et al. In terms of average misclassification error rate, taken over all data sets, SPAN was among the best five methods. In terms of average ranking of misclassification, that is, for each data set ranking the misclassification rates from lowest to highest, SPAN was second only to polyclass logistic regression.

Comparison of the Predictive Accuracy of DNA Array-Based Multigene Classifiers across cDNA Arrays and Affymetrix GeneChips

We examined how well differentially expressed genes and multigene outcome classifiers retain their class-discriminating values when tested on data generated by different transcriptional profiling platforms. RNA from 33 stage I-III breast cancers was hybridized to both Affymetrix GeneChip and Millennium Pharmaceuticals cDNA arrays. Only 30% of all corresponding gene expression measurements on the two platforms had Pearson correlation coefficient r >or= 0.7 when UniGene was used to match probes. There was substantial variation in correlation between different Affymetrix probe sets matched to the same cDNA probe. When cDNA and Affymetrix probes were matched by basic local alignment tool (BLAST) sequence identity, the correlation increased substantially. We identified 182 genes in the Affymetrix and 45 in the cDNA data (including 17 common genes) that accurately separated 91% of cases in supervised hierarchical clustering in each data set. Cross-platform testing of these informative genes resulted in lower clustering accuracy of 45 and 79%, respectively. Several sets of accurate five-gene classifiers were developed on each platform using linear discriminant analysis. The best 100 classifiers showed average misclassification error rate of 2% on the original data that rose to 19.5% when tested on data from the other platform. Random five-gene classifiers showed misclassification error rate of 33%. We conclude that multigene predictors optimized for one platform lose accuracy when applied to data from another platform due to missing genes and sequence differences in probes that result in differing measurements for the same gene.

Parametric vs. Non-Parametric Statistics of Low Resolution Electromagnetic Tomography (LORETA)

This study compared the relative statistical sensitivity of non-parametric and parametric statistics of 3-dimensional current sources as estimated by the EEG inverse solution Low Resolution Electromagnetic Tomography (LORETA). One would expect approximately 5% false positives (classification of a normal as abnormal) at the P < .025 level of probability (two tailed test) and approximately 1% false positives at the P < .005 level. EEG digital samples (2 second intervals sampled 128 Hz, 1 to 2 minutes eyes closed) from 43 normal adult subjects were imported into the Key Institute's LORETA program. We then used the Key Institute's cross-spectrum and the Key Institute's LORETA output files (*.lor) as the 2,394 gray matter pixel representation of 3-dimensional currents at different frequencies. The mean and standard deviation *.lor files were computed for each of the 2,394 gray matter pixels for each of the 43 subjects. Tests of Gaussianity and different transforms were computed in order to best approximate a normal distribution for each frequency and gray matter pixel. The relative sensitivity of parametric vs. non-parametric statistics were compared using a "leave-one-out" cross validation method in which individual normal subjects were withdrawn and then statistically classified as being either normal or abnormal based on the remaining subjects. Log10 transforms approximated Gaussian distribution in the range of 95% to 99% accuracy. Parametric Z score tests at P < .05 cross-validation demonstrated an average misclassification rate of approximately 4.25%, and range over the 2,394 gray matter pixels was 27.66% to 0.11%. At P < .01 parametric Z score cross-validation false positives were 0.26% and ranged from 6.65% to 0% false positives. The non-parametric Key Institute's t-max statistic at P < .05 had an average misclassification error rate of 7.64% and ranged from 43.37% to 0.04% false positives. The nonparametric t-max at P < .01 had an average misclassification rate of 6.67% and ranged from 41.34% to 0% false positives of the 2,394 gray matter pixels for any cross-validated normal subject. In conclusion, adequate approximation to Gaussian distribution and high cross-validation can be achieved by the Key Institute's LORETA programs by using a log10 transform and parametric statistics, and parametric normative comparisons had lower false positive rates than the non-parametric tests.

A Multistage Screening Model for Evaluation and Control of Misclassification Error in the Detection of Hypertension

Hypertension is one of the most important risk factors with respect to coronary heart disease and stroke. The benefits of early detection of hypertension and the subsequent design of follow-up treatment programs are well documented. Consequently, screening programs have been designed to identify subjects as normotensive (normal) or hypertensive (abnormal). In order for these programs to be effective, full participation of the subject population is required. However, such classification programs can incur massive risks of incorrectly classifying subjects as normotensive who are truly hypertensive and incorrectly classifying subjects as hypertensive who are truly normotensive. To date, the only means to reduce these risks of misclassification is to require subjects to make numerous visits for blood pressure measurement before they can be classified. Such requirements reduce the level of participation in screening programs and also delay the identification of subjects who are truly hypertensive, thereby depriving them of the benefits of early detection and immediate follow-up treatment. We propose a multiple-stage screening model that controls for maximum as well as average misclassification error which is used to design and/or evaluate screening programs for hypertension. A multiple-stage screening model not only permits the early detection of subjects who are truly hypertensive, but also requires a much smaller level of participation of subjects, while retaining control of misclassification risks that are comparable to those of screening programs based on numerous visits. We then design multiple-stage screening programs for several different subject populations.

Forest Classification Using Simulated Landsat-D Thematic Mapper Data

SUMMARYThe expected performance of the LANDSAT-D Thematic Mapper (TM) in the forestry context has been evaluated from airborne Multi-Spectral Scanner (MSS) data and compared with the performance of simulated and actual data of the LANDSAT MSS. Specific comparisons are also made to examine the separate effects of the differences in spectral bands, radiometric resolution, and spatial resolution. We find that simulated TM data are significantly better than LANDSAT MSS data for the spectral classification of forest types. If TM imagery is used, the average misclassification error is reduced from 33% to 17%. Most of the improvement is attributable to the number and better wavelength characteristics of the TM spectral bands and, to a lesser extent, the improved TM radiometric resolution. No significant improvement in spectral classification was found as a result of the increased spatial resolution.

Forest Classification Using Simulated Landsat-D Thematic Mapper Data

SUMMARYThe expected performance of the LANDSAT-D Thematic Mapper (TM) in the forestry context has been evaluated from airborne Multi-Spectral Scanner (MSS) data and compared with the performance of simulated and actual data of the LANDSAT MSS. Specific comparisons are also made to examine the separate effects of the differences in spectral bands, radiometric resolution, and spatial resolution. We find that simulated TM data are significantly better than LANDSAT MSS data for the spectral classification of forest types. If TM imagery is used, the average misclassification error is reduced from 33% to 17%. Most of the improvement is attributable to the number and better wavelength characteristics of the TM spectral bands and, to a lesser extent, the improved TM radiometric resolution. No significant improvement in spectral classification was found as a result of the increased spatial resolution.