Enhanced Classification Method Research Articles

Enhanced Classification Method for Phishing Emails Detection

Emails are currently the main communication method worldwide as it proven in its efficiency. Phishing emails in the other hand is one of the major threats which results in significant losses, estimated at billions of dollars. Phishing emails is a more dynamic problem, a struggle between the phishers and defenders where the phishers have more flexibility in manipulating the emails features and evading the anti-phishing techniques. Many solutions have been proposed to mitigate the phishing emails impact on the targeted sectors, but none have achieved 100% detection and accuracy. As phishing techniques are evolving, the solutions need to be evolved and generalized in order to mitigate as much as possible. This article presents a new emergent classification model based on hybrid feature selection method that combines two common feature selection methods, Information Gain and Genetic Algorithm that keep only significant and high-quality features in the final classifier. The Proposed hybrid approach achieved 98.9% accuracy rate against phishing emails dataset comprising 8266 instances and results depict enhancement by almost 4%. Furthermore, the presented technique has contributed to reducing the search space by reducing the number of selected features.

Improving Image-Based Plant Disease Classification With Generative Adversarial Network Under Limited Training Set.

Traditionally, plant disease recognition has mainly been done visually by human. It is often biased, time-consuming, and laborious. Machine learning methods based on plant leave images have been proposed to improve the disease recognition process. Convolutional neural networks (CNNs) have been adopted and proven to be very effective. Despite the good classification accuracy achieved by CNNs, the issue of limited training data remains. In most cases, the training dataset is often small due to significant effort in data collection and annotation. In this case, CNN methods tend to have the overfitting problem. In this paper, Wasserstein generative adversarial network with gradient penalty (WGAN-GP) is combined with label smoothing regularization (LSR) to improve the prediction accuracy and address the overfitting problem under limited training data. Experiments show that the proposed WGAN-GP enhanced classification method can improve the overall classification accuracy of plant diseases by 24.4% as compared to 20.2% using classic data augmentation and 22% using synthetic samples without LSR.

A novel approach for scene classification from remote sensing images using deep learning methods

ABSTRACT Deep learning plays a major task in classification of unsupervised data, which utilises network enhanced learning. This technique proves to be powerful in remote sensing (RS) field for spatial data classification. In the existing environment, huge amounts of data from the earth observation satellites known as satellite images time series (SITS) are gathered, which can be utilised for observing the areas related to geography over through time. In this proposed model the time series model utilised is based on geography. There exists a challenge on how these types of information can be analysed in the field of remote sensing. Notable, techniques related to deep learning substantiated in dealing with remote sensing usually for classification of scene . In this paper, we propose an enhanced classification method involving Recurrent Neural Network (RNN) along with Random forest (RF) for land classification using satellite images, which are publicly available for various research purposes. We utilised spatial data gathered from the satellite images (i.e. time series). Our experimental classification is based on pixel and object-based classification. The attained analysis illustrates that the proposed model outperforms the other present day remote sensing classification techniques by producing 87% target accuracy of classification scene from satellite images.

Limb-position robust classification of myoelectric signals for prosthesis control using sparse representations.

The fundamental objective in non-invasive myoelectric prosthesis control is to determine the user's intended movements from corresponding skin-surface recorded electromyographic (sEMG) activation signals as quickly and accurately as possible. Linear Discriminant Analysis (LDA) has emerged as the de facto standard for real-time movement classification due to its ease of use, calculation speed, and remarkable classification accuracy under controlled training conditions. However, performance of cluster-based methods like LDA for sEMG pattern recognition degrades significantly when real-world testing conditions do not resemble the trained conditions, limiting the utility of myoelectrically controlled prosthesis devices. We propose an enhanced classification method that is more robust to generic deviations from training conditions by constructing sparse representations of the input data dictionary comprised of sEMG time-frequency features. We apply our method in the context of upper-limb position changes to demonstrate pattern recognition robustness and improvement over LDA across discrete positions not explicitly trained. For single position training we report an accuracy improvement in untrained positions of 7.95%, p ≪ .001, in addition to significant accuracy improvements across all multiposition training conditions, p <; .001.

Monitoring the state of a large boreal forest region in eastern Canada through the use of multitemporal classified satellite imagery

Multitemporal classification of Landsat imagery was used to measure and monitor the state of the forest over a large area (11.6 million ha) of boreal forest in eastern Canada using four criteria for a 20 year period (1985–2005). The Enhancement-Classification Method was used in this study. Forty-eight thematic classes based on Canada's National Forest Inventory were identified, then grouped into 13 indicators, and reorganized within four main criteria: (i) forest versus nonforest land cover, (ii) forest development stage, (iii) forest cover type, and (iv) forest cover density. Validation based on 2973 high-resolution geo-referenced digital aerial colour photos of the 2005 classified images showed an overall accuracy of the four criteria of 83%, 68%, 58%, and 62%, respectively. The change in each indicator between 1985 and 2005 could be summarized as: (i) a decrease in productive forest area of 0.4% (approx. 43 000 ha); (ii) a 4.6% decrease in mature stand area, with a concomitant increase in areas classified as vegetated (1.3%) and regenerated (3.4%); (iii) concentration of harvesting pressure on coniferous and mixed stands with respective reductions of 8.2% and 0.8%, due to their conversion to deciduous stands; and (iv) an increase in low-density stands (3.1%) and a decrease in high-density stands (8.3%). These results demonstrate that medium-resolution (30 m) remote sensing tools can be used both to monitor the state of the boreal forest and to produce key indicators, which were extracted from the multidate Landsat satellite imagery.

An enhanced classification method comprising a genetic algorithm, rough set theory and a modified PBMF-index function

This study proposes a method, designated as the GRP-index method, for the classification of continuous value datasets in which the instances do not provide any class information and may be imprecise and uncertain. The proposed method discretizes the values of the individual attributes within the dataset and achieves both the optimal number of clusters and the optimal classification accuracy. The proposed method consists of a genetic algorithm (GA) and an FRP-index method. In the FRP-index method, the conditional and decision attribute values of the instances in the dataset are fuzzified and discretized using the Fuzzy C-means (FCM) method in accordance with the cluster vectors given by the GA specifying the number of clusters per attribute. Rough set (RS) theory is then applied to determine the lower and upper approximate sets associated with each cluster of the decision attribute. The accuracy of approximation of each cluster of the decision attribute is then computed as the cardinality ratio of the lower approximate sets to the upper approximate sets. Finally, the centroids of the lower approximate sets associated with each cluster of the decision attribute are determined by computing the mean conditional and decision attribute values of all the instances within the corresponding sets. The cluster centroids and accuracy of approximation are then processed by a modified form of the PBMF-index function, designated as the RP-index function, in order to determine the optimality of the discretization/classification results. In the event that the termination criteria are not satisfied, the GA modifies the initial population of cluster vectors and the FCM, RS and RP-index function procedures are repeated. The entire process is repeated iteratively until the termination criteria are satisfied. The maximum value of the RP cluster validity index is then identified, and the corresponding cluster vector is taken as the optimal classification result. The validity of the proposed approach is confirmed by cross validation, and by comparing the classification results obtained for a typical stock market dataset with those obtained by non-supervised and pseudo-supervised classification methods. The results show that the proposed GRP-index method not only has a better discretization performance than the considered methods, but also achieves a better accuracy of approximation, and therefore provides a more reliable basis for the extraction of decision-making rules.

Mapping lichen in a caribou habitat of Northern Quebec, Canada, using an enhancement_classification method and spectral mixture analysis

Studies of caribou herds in northern regions are important to better understand population dynamics and define wildlife management strategies. Lichen is a primary food source for caribou and is a good indicator of caribou herd activity because of its sensitivity to overgrazing and overtrampling, its widespread distribution over northern areas, and its influence on herd demography. In this paper, we used Landsat TM imagery for mapping lichen in the summer range of the George River caribou herd in northern Quebec, Canada. Results from the enhancement_classification method (ECM) and from spectral mixture analysis (SMA) were evaluated for their suitability to characterize lichen landcover and for their potential to be applied over large territories. ECM and SMA are assessed individually, and also for potential synergistic use. ECM is based on guided unsupervised classification of enhanced satellite images. Validation based on 3536 pixels from a relatively smaller number of field sites (20) showed an overall accuracy of 74.5% (kappa=0.70) for 10 classes and good discrimination between lichen and non_lichen classes, though we interpret these results with caution due to spatial autocorrelation and non_random sampling within field sites. However, discrimination amongst different lichen classes using ECM was more problematic. SMA derives the proportion of individual scene components at sub_pixel scales. This method provided good results in characterizing variations in lichen abundance validated against field observations, and provided additional and new information not provided by ECM which is important since the abundance of lichen as a primary food source is a key indicator of migration and demographic patterns essential for effective wildlife management. We concluded that the ECM and SMA methods are appropriate for different aspects of lichen mapping. ECM provided good discrimination between lichen and non_lichen classes whereas SMA provided additional lichen information not available by classification yet critical to the environmental application, which is also appropriate for application over much larger areas and in spatio_temporal studies. A synergistic use of SMA and ECM is therefore recommended for future research.

Large area forest classification and biophysical parameter estimation using the 5-Scale canopy reflectance model in Multiple-Forward-Mode

The Multiple-Forward-Mode approach to running the 5-Scale geometric-optical reflectance model (MFM-5-Scale) provides an inversion modeling capability for powerful but non-invertible models, and yields both landcover classification and forest biophysical–structural information that is important in global change projects such as BOREAS. Unlike regular forward mode, MFM-5-Scale does not require exact physical stand descriptors, but instead requires only input ranges and model increments that are easily obtained, with results determined by matching satellite image data with modeled reflectance values as stored in a look-up table. In this work, MFM-5-Scale was applied to a mosaic of seven multi-year (1991–1998) Landsat TM scenes covering the BOREAS region in western Canada, with results validated against field data and also compared with the Enhancement Classification Method (ECM), a highly accurate yet subjective and labour intensive approach which involves considerable user judgement and expertise. The goal was to approach ECM accuracy using MFM-5-Scale, but without the subjectivity of ECM. Classifications of 13 forest classes including deciduous and species-specific coniferous classes subdivided into low, medium and high crown densities as well as a mixed forest class were evaluated against 136 field checked validation sites. Overall classification accuracies were 91% for ECM and 85% for MFM-5-Scale. Additional assessments of agreement between the MFM-5-Scale and ECM products were also performed over larger sample areas. MFM-5-Scale classification of a full set of 25 forest and non-forest classes adhering to Global Observation of Forest Cover (GOFC) specifications was in 80% agreement with the ECM product ( n=11,442). This is viewed as a significant result given the ambitious and detailed nature of the classes considered (21 forest classes comprising 12 species and density specific conifer and deciduous classes, four mixed forest classes and five burnt forest classes, as well as four non-forest classes). The two products were also in 94% agreement for 6 forest density classes ( n=3730). MFM-5-Scale biophysical analysis of 63 BOREAS plots showed LAI was estimated within ±0.53 LAI compared with ground-based TRAC LAI validation data (biophysical information is not provided by ECM). These results represent significant progress towards defining an operational landcover and biophysical estimation approach given the objective, semi-automated nature of MFM-5-Scale physical scene modeling compared to subjective, user-driven methods such as ECM.

Land cover from multiple thematic mapper scenes using a new enhancement‐classification methodology

The purpose of this paper is to test a methodology for extracting land cover information from high‐resolution satellite images over large areas. The use of multiple scenes brings additional complications which are related to atmospheric, phenological, spectral, and classification legend aspects of the data set. The approach is based on a supervised digital mosaicking of the input scenes and on a guided unsupervised classification of the mosaic. The goal of the development was to produce a methodology which is robust, minimally sensitive to the biases of the analyst, and capable of extracting all land cover type information contained in the satellite data. The mosaicking procedure aims to achieve radiometric consistency across the mosaic for key cover types. The classification procedure, named enhancement‐classification method (ECM), operates on three (or less) input channels and produces a classification in which virtually the entire relevant spectral content has been extracted. The above procedures were employed to produce a land cover classification of the BOREAS transect from six Landsat thematic mapper images. Initial assessment indicated that the classification accuracy of the mosaics varies with cover type (44–82% for forest cover types). The variability appears related to residual radiometric effects, lack of unique spectral land cover signatures, and vegetation phenology. Thus nonremote sensing (or multitemporal) information is likely to be required for consistent accuracies. While the two‐step procedure could be successfully automated under some conditions, the input of an analyst will remain essential until more experience is obtained, especially with classification across multiple scenes.