Supervised Classification Techniques Research Articles

CLUSTERING-BASED NETWORK INTRUSION DETECTION

Recently data mining methods have gained importance in addressing network security issues, including network intrusion detection — a challenging task in network security. Intrusion detection systems aim to identify attacks with a high detection rate and a low false alarm rate. Classification-based data mining models for intrusion detection are often ineffective in dealing with dynamic changes in intrusion patterns and characteristics. Consequently, unsupervised learning methods have been given a closer look for network intrusion detection. We investigate multiple centroid-based unsupervised clustering algorithms for intrusion detection, and propose a simple yet effective self-labeling heuristic for detecting attack and normal clusters of network traffic audit data. The clustering algorithms investigated include, k-means, Mixture-Of-Spherical Gaussians, Self-Organizing Map, and Neural-Gas. The network traffic datasets provided by the DARPA 1998 offline intrusion detection project are used in our empirical investigation, which demonstrates the feasibility and promise of unsupervised learning methods for network intrusion detection. In addition, a comparative analysis shows the advantage of clustering-based methods over supervised classification techniques in identifying new or unseen attack types.

Asteroid taxonomic classification in the Gaia photometric system

Aims. We evaluate the prospects of performing automatic taxonomic classification of asteroids in a proposed broad and medium band photometric system of Gaia .Methods. The study is based on asteroid colors from the Eight-Color Asteroid Survey (ECAS) and CCD spectra from the Small Mainbelt Asteroid Spectral Survey II (SMASSII). The success of the Gaia photometric systems for taxonomic classification is evaluated using supervised classification techniques and mean taxonomic class spectra in the Tholen and Bus&Binzel taxonomic systems. Our supervised classification method is based on rms differences between individual asteroid spectra and taxonomic mean spectra and provides probability estimates of membership in all taxonomic classes.Results. We find that both photometric systems of Gaia are able to discriminate between all of the twelve Tholen asteroid classes for noise-free data. The medium band system is able to discriminate between the majority of the 26 SMASSII asteroid classes in case of high quality photometric data. For both the Tholen and Bus & Binzel taxonomies we find that about 25% of the asteroids are spectrally more similar to another taxonomic class in a best-fit sense, though the differences within the three major complexes (C, S and X) are 1–10%.Conclusions. Among the two main existing taxonomies, the Gaia photometric system is found to be best suited for the Bus & Binzel taxonomy. The medium band system is the preferred system for all but the faintest objects. The classification method employed here results in more concentrated taxonomic class domains in principal component space, and mean taxonomic spectra that are formed from less divergent class members, than the case for the nominal classification systems. It provides statistical probability estimates for class memberships and naturally reflects the fact that asteroid spectral shapes form a continuum in principal component space.

A Fuzzy Min-Max Neural Network Classifier With Compensatory Neuron Architecture

This paper proposes a fuzzy min-max neural network classifier with compensatory neurons (FMCNs). FMCN uses hyperbox fuzzy sets to represent the pattern classes. It is a supervised classification technique with new compensatory neuron architecture. The concept of compensatory neuron is inspired from the reflex system of human brain which takes over the control in hazardous conditions. Compensatory neurons (CNs) imitate this behavior by getting activated whenever a test sample falls in the overlapped regions amongst different classes. These neurons are capable to handle the hyperbox overlap and containment more efficiently. Simpson used contraction process based on the principle of minimal disturbance, to solve the problem of hyperbox overlaps. FMCN eliminates use of this process since it is found to be erroneous. FMCN is capable to learn the data online in a single pass through with reduced classification and gradation errors. One of the good features of FMCN is that its performance is less dependent on the initialization of expansion coefficient, i.e., maximum hyperbox size. The paper demonstrates the performance of FMCN by comparing it with fuzzy min-max neural network (FMNN) classifier and general fuzzy min-max neural network (GFMN) classifier, using several examples.

Landscape pattern and eco-hydrological characteristics at the upstream of Minjiang River, China

Based on three scenes of Landsat 5 Thematic Mapper (TM) satellite images acquired on June 26, 1994, 12 land cover types were identified by the supervised classification techniques. The precipitation, runoff, and normalized difference vegetation index (NDVI) data of six catchments were accumulated from April to September in 1992, 1993, and 1995. A new eco-hydrological index, expressed by the difference between precipitation and runoff divided by the product of precipitation and NDVI, was used in this study to represent the eco-hydrological functions of different catchments. The results were: (1) The selected six catchments at the upstream of Minjiang River, China were different in landscape patterns in terms of landscape type and cover. There were higher contagion, lower edge density and diversity index in the Shouxi catchments and lower contagion, higher edge density and diversity index in the Zagunao catchments. (2) Eco-hydrological indexes had remarkable differences among different catchments. The highest eco-hydrological index was found in the Shouxi catchments, which indicated higher precipitation holding capacity of vegetation therein. While the lower eco-hydrological index was found in the Zagunao catchments, which indicated its lower precipitation holding capacity of vegetation. (3) High correlation was detected between the landscape indexes and eco-hydrological indexes. Eco-hydrological index was positively correlated with landscape contagion in contrast with the negative correlation with landscape diversity and edge density.

Modelo de clasificación dinámico de arritmias cardíacas mediante aprendizaje de máquina con interfaz a usuario

Este trabajo presenta el desarrollo de un modelo de clasificación de arritmias cardiacas, basado en la duración de las ondas de registros electrocardiográficos, utilizando técnicas de clasificación supervisada. Los clasificadores desarrollados se basaron en técnicas de aprendizaje de máquina, Se implementaron metaclasificadores para los diferentes clasificadores. El desempeño de los diferentes clasificadores se probó aplicando criterios objetivos de comparación, y se obtuvieron resultados prometedores. Para presentar los resultados al usuario final, se diseñó una interfaz para ser usada por el especialista como ayuda a la clasificación de arritmias.

Land Use – Land Cover Conversion, Regeneration and Degradation in the High Elevation Bolivian Andes

Regional land-cover change affects biodiversity, hydrology, and biogeochemical cycles at local, watershed, and landscape scales. Developing countries are experiencing rapid land cover change, but assessment is often restricted by limited financial resources, accessibility, and historical data. The assessment of regional land cover patterns is often the first step in developing conservation and management plans. This study used remotely sensed land cover and topographic data (Landsat and Shuttle Radar Topography Mission), supervised classification techniques, and spectral mixture analysis to characterize current landscape patterns and quantify land cover change from 1985 to 2003 in the Altiplano (2535–4671 m) and Intermediate Valley (Mountain) (1491–4623 m) physiographic zones in the Southeastern Bolivian Andes. Current land cover was mapped into six classes with an overall accuracy of 88% using traditional classification techniques and limited field data. The land cover change analysis showed that extensive deforestation, desertification, and agricultural expansion at a regional scale occurred in the last 20 years (17.3% of the Mountain Zone and 7.2% of the Altiplano). Spectral mixture analysis (SMA) indicated that communal rangeland degradation has also occurred, with increases in soil and non-photosynthetic vegetation fractions in most cover classes. SMA also identified local areas with intensive management activities that are changing differently from the overall region (e.g., localized areas of increased green vegetation). This indicates that actions of local communities, governments, and environmental managers can moderate the potentially severe future changes implied by the results of this study.

Discovering Clinical Biomarkers of Ionizing Radiation Exposure with Serum Proteomic Analysis

In this study, we sought to explore the merit of proteomic profiling strategies in patients with cancer before and during radiotherapy in an effort to discover clinical biomarkers of radiation exposure. Patients with a diagnosis of cancer provided informed consent for enrollment on a study permitting the collection of serum immediately before and during a course of radiation therapy. High-resolution surface-enhanced laser desorption and ionization-time of flight (SELDI-TOF) mass spectrometry (MS) was used to generate high-throughput proteomic profiles of unfractionated serum samples using an immobilized metal ion-affinity chromatography nickel-affinity chip surface. Resultant proteomic profiles were analyzed for unique biomarker signatures using supervised classification techniques. MS-based protein identification was then done on pooled sera in an effort to begin to identify specific protein fragments that are altered with radiation exposure. Sixty-eight patients with a wide range of diagnoses and radiation treatment plans provided serum samples both before and during ionizing radiation exposure. Computer-based analyses of the SELDI protein spectra could distinguish unexposed from radiation-exposed patient samples with 91% to 100% sensitivity and 97% to 100% specificity using various classifier models. The method also showed an ability to distinguish high from low dose-volume levels of exposure with a sensitivity of 83% to 100% and specificity of 91% to 100%. Using direct identity techniques of albumin-bound peptides, known to underpin the SELDI-TOF fingerprints, 23 protein fragments/peptides were uniquely detected in the radiation exposure group, including an interleukin-6 precursor protein. The composition of proteins in serum seems to change with ionizing radiation exposure. Proteomic analysis for the discovery of clinical biomarkers of radiation exposure warrants further study.

Lithological mapping in the Eastern Desert of Egypt, the Barramiya area, using Landsat thematic mapper (TM)

The Barramiya area, Central Eastern Desert, Egypt, comprises a variety of Late Proterozoic rocks including serpentinites, island-arc assemblage of metasediments and metavolcanics, and syntectonic granitoids. Lithological mapping in the Barramiya area is carried out by using Landsat thematic mapper (TM) image enhancement techniques, including RGB band ratio and supervised classification images. The significant spectral characteristics of serpentinites have been used previously by other authors to map these rocks using different band ratio images. In the current study, band ratio images (5/3, 5/1, 7/5), and (7/5, 5/4, 3/1) together with supervised classification techniques are used and proved effective for lithological mapping of the serpentinite rocks. Integrating previous knowledge lithological mapping studies of the study area and our visual interpretations of the different enhanced Landsat images, we mapped, in detail, the distribution of the Pan-African serpentinites in the Barramiya area.

Classification of dopamine antagonists using functional feature hypothesis and topological descriptors

The designing of selective dopamine antagonists for their own subreceptors can be useful in individual therapy of various neuropsychiatric disorders. Three-dimensional pharmacophore hypothesis and two-dimensional topological descriptors were used to investigate and compare different classes of dopamine antagonists. The structurally diverse D 3 and D 4 antagonists above preclinical trials were selected to map common structural features of highly selective and efficacious antagonists. The generated pharmacophore hypotheses were successfully employed as discriminative probe for database screening. To filter out the false positive from screening hits, the classification models by two-dimensional topological descriptors were built. Molconn-Z and BCUT topological descriptors were employed to develop a classification model for 1328 dopamine antagonists from MDDR database. The soft independent modeling of class analogy and artificial neural network, two supervised classification techniques, successfully classified D 1, D 3, and D 4 antagonists at the average of 80% rates into their own active classes. The mean classification rates for D 2 antagonists were obtained to 60% due to insufficient selective D 2 antagonists. In this paper, we report the validity of our models generated using functional feature hypotheses and topological descriptors. The combining both of classification using functional feature hypotheses and topological descriptors would be a useful tool to predict selective antagonists.

Word segmentation of handwritten text using supervised classification techniques

Recent work on extracting features of gaps in handwritten text allows a classification of these gaps into inter-word and intra-word classes using suitable classification techniques. In this paper, we first analyse the features of the gaps using mutual information. We then investigate the underlying data distribution by using visualisation methods. These suggest that a complicated structure exists, which makes them difficult to be separated into two distinct classes. We apply five different supervised classification algorithms from the machine learning field on both the original dataset and a dataset with the best features selected using mutual information. Moreover, we improve the classification result with the aid of a set of feature variables of strokes preceding and following each gap. The classifiers are compared by employing McNemar's test. We find that SVMs and MLPs outperform the other classifiers and that preprocessing to select features works well. The best classification result attained suggests that the technique we employ is particularly suitable for digital ink manipulation at the level of words.

Class Discovery in Galaxy Classification

In recent years, automated, supervised classification techniques have been fruitfully applied to labeling and organizing large astronomical databases. These methods require off-line classifier training, based on labeled examples from each of the (known) object classes. In practice, only a small batch of labeled examples, hand-labeled by a human expert, may be available for training. Moreover, there may be no labeled examples for some classes present in the data; i.e., the database may contain several unknown classes. Unknown classes may be present because of (1) uncertainty in or lack of knowledge of the measurement process, (2) an inability to adequately survey a massive database to assess its content (classes), and/or (3) an incomplete scientific hypothesis. In recent work, the question of new class discovery in mixed labeled/unlabeled data was formally posed, with a proposed solution based on mixture models. In this work we investigate this approach, propose a competing technique suitable for class discovery in neural networks, and evaluate methods for both classification and class discovery in several astronomical data sets. Our results demonstrate up to a 57% reduction in classification error compared to a standard neural network classifier that uses only labeled data.

A Search Coding Method and Its Application in Supervised Classification

Supervised classification based on error-correcting output codes (ECOC) is a new research direction to improve the generalization of classifiers, yet there is no general method to construct ECOC for any number of classes. This paper analyzes the properties of ECOC and presents a search coding method which corresponds to codewords with integers and gets a satisfied output code through searching an integer range in sequence. It then describes the supervised classification technique based on the search coding method. By applying the search coding method to naive-Bayes algorithm and BP neural networks, experimental results show that the method is an effective

Molecular Diagnosis

We discuss supervised classification techniques applied to medical diagnosis based on gene expression profiles. Our focus lies on strategies of adaptive model selection to avoid overfitting in high-dimensional spaces. We introduce likelihood-based methods, classification trees, support vector machines and regularized binary regression. For regularization by dimension reduction, we describe feature selection methods: feature filtering, feature shrinkage and wrapper approaches. In small sample-size situations efficient methods of data re-use are needed to assess the predictive power of a model. We discuss two issues in using cross-validation: the difference between in-loop and out-of-loop feature selection, and estimating model parameters in nested-loop cross-validation. Gene selection does not reduce the dimensionality of the model. Tuning parameters enable adaptive model selection. The feature selection bias is a common pitfall in performance evaluation. Model selection and performance evaluation can be combined by nested-loop cross-validation. Classification of microarrays is prone to overfitting. A rigorous and unbiased assessment of the predictive power of the model is a must.

GENE SELECTION FOR CANCER CLASSIFICATION USING WRAPPER APPROACHES

Despite the fact that cancer classification has considerably improved, nowadays a general method that classifies known types of cancer has not yet been developed. In this work, we propose the use of supervised classification techniques, coupled with feature subset selection algorithms, to automatically perform this classification in gene expression datasets. Due to the large number of features of gene expression datasets, the search of a highly accurate combination of features is done by means of the new Estimation of Distribution Algorithms paradigm. In order to assess the accuracy level of the proposed approach, the naïve-Bayes classification algorithm is employed in a wrapper form. Promising results are achieved, in addition to a considerable reduction in the number of genes. Stating the optimal selection of genes as a search task, an automatic and robust choice in the genes finally selected is performed, in contrast to previous works that research the same types of problems.

Mapping vegetation in a heterogeneous mountain rangeland using landsat data: an alternative method to define and classify land-cover units

Three major problems are faced when mapping natural vegetation with mid-resolution satellite images using conventional supervised classification techniques: defining the adequate hierarchical level for mapping; defining discrete land cover units discernible by the satellite; and selecting representative training sites. In order to solve these problems, we developed an approach based on the: (1) definition of ecologically meaningful units as mosaics or repetitive combinations of structural types, (2) utilization of spectral information (indirectly) to define the units, (3) exploration of two alternative methods to classify the units once they are defined: the traditional, Maximum Likelihood method, which was enhanced by analyzing objective ways of selecting the best training sites, and an alternative method using Discriminant Functions directly obtained from the statistical analysis of signatures. The study was carried out in a heterogeneous mountain rangeland in central Argentina using Landsat data and 251 field sampling sites. On the basis of our analysis combining terrain information (a matrix of 251 stands×14 land cover attributes) and satellite data (a matrix of 251 stands×8 bands), we defined 8 land cover units (mosaics of structural types) for mapping, emphasizing the structural types which had stronger effects on reflectance. The comparison through field validation of both methods for mapping units showed that classification based on Discriminant Functions produced better results than the traditional Maximum Likelihood method (accuracy of 86% vs. 78%).

QUANTIFYING ADULTERATION IN ROAST COFFEE POWDERS BY DIGITAL IMAGE PROCESSING

Abstract A new method based on digital image processing used to quantify adulteration in roast and ground coffee beans is presented in this study. Pure arabica coffee and mixtures of coffee husks and straw, maize, brown sugar and soybean were produced in our laboratory as investigation materials. Red/Green/Blue (RGB) color composites, magnified twelve times, were generated using a Charge Coupled Device (CCD) camera connected to a stereo microscope and a personal computer with an image processing software package. The percent areas of the contaminants in each image were calculated by the Maximum Likelihood supervised classification technique. Best‐fit equations relating weight percentage (g.kg‐1) and the percent areas were obtained for each coffee contaminant. To test the method, 247 coffee samples of different amounts and types of adulterants were analyzed in the laboratory. The results showed that the new method developed can analyze precisely and quickly a large number of ground coffee powders.

Using routine meteorological data to derive sky conditions

Abstract. Sky condition is a matter of interest for public and weather predictors as part of weather analyses. In this study, we apply a method that uses total solar radiation and other meteorological data recorded by an automatic station for deriving an estimation of the sky condition. The impetus of this work is the intention of the Catalan Meteorological Service (SMC) to provide the public with real-time information about the sky condition. The methodology for deriving sky conditions from meteorological records is based on a supervised classification technique called maximum likelihood method. In this technique we first need to define features which are derived from measured variables. Second, we must decide which sky conditions are intended to be distinguished. Some analyses have led us to use four sky conditions: (a) cloudless or almost cloudless sky, (b) scattered clouds, (c) mostly cloudy – high clouds, (d) overcast – low clouds. An additional case, which may be treated separately, corresponds to precipitation (rain or snow). The main features for estimating sky conditions are, as expected, solar radiation and its temporal variability. The accuracy of this method of guessing sky conditions compared with human observations is around 70% when applied to four sites in Catalonia (NE Iberian Peninsula). The agreement increases if we take into account the uncertainty both in the automatic classifier and in visual observations.Key words. Meteorological and atmospheric dynamics (instruments and techniques; radiative processes) – Atmospheric composition and structure (cloud physics and chemistry)

Analysis on the spatial distribution variation characteristic of urban heat environmental quality and its mechanism

Urban heat environmental quality (UHEQ) is affected by the interacting of weather condition and underlying surface framework of urban area. In the last two decades, many researchers from domestic and overseas have studied many problems at the aspect of urban heat environment such as urban heat islands, urban air temperature and their relation with urban land cover, city population, air pollution etc. In the recent years, Hangzhou, acting as a center city of Zhejiang Province in China, its urbanization quantum and quantity have both changed greatly, in particular, representing as business affairs building, resident real property and all kinds of specialty market having arisen in built-up zone. Based on Landsat TM images data in 1991 and 1999, urban underlying surface temperature value and Normalized Difference Vegetation Index (NDVI) were calculated using image interpreting and supervised classification technique by remote sensing software ERDAS image 8.4. The relation model between urban underlying surface temperature (UUST) and urban air temperature was setup according to the certain correlation pattern. Reference to the relational standard of assessing human comfort and other meteorology data of Hangzhou City in summer, the spatial distribution characteristic and the spatial variation degree of human comfort of heat environmental quality are estimated and mapped on a middle scale, that is, in six districts of Hangzhou City. Then the paper reveals the main characteristics of spatial variation from 1991 to 1999. Lastly, the change mechanism is analyzed and discussed from the viewpoint of city planning, construction and environmental protection.

Classification of SAR images using a general and tractable multiplicative model

Among the frameworks for Synthetic Aperture Radar (SAR) image modelling and analysis, the multiplicative model is very accurate and successful. It is based on the assumption that the observed random field is the result of the product of two independent and unobserved random fields: X and Y. The random field X models the terrain backscatter and, thus, depends only on the type of area to which each pixel belongs. The random field Y takes into account that SAR images are the result of a coherent imaging system that produces the well-known phenomenon called speckle noise, and that they are generated by performing an average of n statistically independent images (looks) in order to reduce the noise effect. There are various ways of modelling the random field X; recently the Γ−1/2(α, γ) distribution was proposed. This, with the usual Γ1/2(n, n) distribution for the amplitude speckle, resulted in a new distribution for the return: the (α, γ, n) law. The parameters α and γ depend only on the ground truth, and n is the number of looks. The advantage of this distribution over the ones used in the past is that it models very well extremely heterogeneous areas like cities, as well as moderately heterogeneous areas like forests and homogeneous areas like pastures. As the ground data can be characterized by the parameters α and γ, their estimation in each pixel generates parameter maps that can be used as the input for classification methods. In this work, moment estimators are used on simulated and on real SAR images and, then, a supervised classification technique (Gaussian maximum likelihood) is performed and evaluated. Excellent classification results are obtained.

Remote sensing: A technology for assessment of sugarcane crop acreage and yield

A study was conducted for identification and acreage estimation of sugarcane crop in KCP Sugar factory zones of Vuyyuru and Lakshmipuram using satellite remote sensing data during cane seasons of 1997 to 2000. The satellite data was analyzed using supervised classification techniques for acreage estimation and Normalized Difference Vegetation Index (NDVI) was employed for condition assessment and yield estimation. The acreage estimation was made possible after 100 days of planting of sugarcane and the cane crop is identifiable on satellite data after April month in the state of Andhra Pradesh. The relationship between yields of sugarcane with NDVI shown a positive correlation with correlation coefficient of 0.84. However, these relationships needs to be developed for each factory-wise in order to represent the local soil, climatic and biotic factors.