Study on ensemble classification methods towards spam filtering

Recently, many scholars make use of fusion of filters to enhance the performance of spam filtering. In the past several years, a lot of effort has been devoted to different ensemble methods to achieve better performance. In reality, how to select appropriate ensemble methods towards spam filtering is an unsolved problem. In this paper, we investigate this problem through designing a framework to compare the performances among various ensemble methods. It is helpful for researchers to fight spam email more effectively in applied systems. The experimental results indicate that online based methods perform well on accuracy, while the off-line batch methods are evidently influenced by the size of data set. When a large data set is involved, the performance of off-line batch methods is not at par with online methods, and in the framework of online methods, the performance of parallel ensemble is better when using complex filters only.

Spam filtering for network traffic security on a multi-core environment

This paper presents an innovative fusion-based multi-classifier e-mail classification on a ubiquitous multicore architecture. Many previous approaches used text-based single classifiers to identify spam messages from a large e-mail corpus with some amount of false positive tradeoffs. Researchers are trying to prevent false positive in their filtering methods, but so far none of the current research has claimed zero false positive results. In e-mail classification false positive can potentially cause serious problems for the user. In this paper, we use fusion-based multi-classifier classification technique in a multi-core framework. By running each classifier process in parallel within their dedicated core, we greatly improve the performance of our multi-classifier-based filtering system in terms of running time, false positive rate, and filtering accuracy. Our proposed architecture also provides a safeguard of user mailbox from different malicious attacks. Our experimental results show that we achieved an average of 30% speedup at an average cost of 1.4 ms. We also reduced the instances of false positives, which are one of the key challenges in a spam filtering system, and increases e-mail classification accuracy substantially compared with single classification techniques.

An approach to filtering RFID data streams

RFID is gaining significant thrust as the preferred choice of automatic identification and data collection system. However, there are various data processing and management problems such as missed readings and duplicate readings which hinder wide scale adoption of RFID systems. To this end we propose an approach that filters the captured data including both noise removal and duplicate elimination. Experimental results demonstrate that the proposed approach improves missed data restoration process when compared with the existing method.

Minimizing the drawbacks of grey-list analyser in synthesis based spam filtering

In the last decade, the rapid growth of the Internet and email, there has been a dramatic growth in spam. Spam is commonly defined as unsolicited email messages and protecting email from the infiltration of spam is an important research issue. Classifications algorithms have been successfully used to filter spam, but with a certain amount of false positive trade-offs, which is unacceptable to users sometimes. This paper presents an approach of email classification to overcome the burden of analyzing technique of GL (grey list) analyzer as further refinements of synthesis based email classification technique. In this approach, we introduce a “majority voting grey list (MVGL)” analyzing technique which will analyze the GL emails by using the majority voting (MV) algorithm. We have presented two different variations of the MV system, one is simple MV (SMV) and other is the Ranked MV (RMV). Our empirical evidence proofs the improvements of this approach compared to existing GL analyzer [7].

MVGL analyser for multi-classifier based spam filtering system

In the last decade, the rapid growth of the Internet and email, there has been a dramatic growth in spam. Spam is commonly defined as unsolicited email messages and protecting email from the infiltration of spam is an important research issue. Classifications algorithms have been successfully used to filter spam, but with a certain amount of false positive trade-offs, which is unacceptable to users sometimes. This paper presents an approach to overcome the burden of GL (grey list) analyzer as further refinements to our multi-classifier based classification model (Islam, M. and W. Zhou 2007). In this approach, we introduce a ldquomajority voting grey list (MVGL)rdquo analyzing technique which will analyze the generated GL emails by using the majority voting (MV) algorithm. We have presented two different variations of the MV system, one is simple MV (SMV) and other is the ranked MV (RMV). Our empirical evidence proofs the improvements of this approach compared to the existing GL analyzer of multi-classifier based spam filtering process.

An approach to filtering duplicate RFID data streams

In a system where distributed network of Radio Frequency Identification (RFID) readers are used to collaboratively collect data from tagged objects, a scheme that detects and eliminates redundant data streams is required. To address this problem, we propose an approach that is based on Bloom filter to detect duplicate readings and filter redundant RFID data streams. We have evaluated the performance of the proposed approach and compared it with existing approaches. The experimental results demonstrate that the proposed approach provides superior performance as compared to the baseline approaches.

Accuracy of ARGOS locations of pinnipeds at-sea estimated using Fastloc GPS

Background: ARGOS satellite telemetry is one of the most widely used methods to track the movements of free-ranging marine and terrestrial animals and is fundamental to studies of foraging ecology, migratory behavior and habitat-use. ARGOS location estimates do not include complete error estimations, and for many marine organisms, the most commonly acquired locations (Location Class 0, A, B, or Z) are provided with no declared error estimate.
Methodology/Principal Findings: We compared the accuracy of ARGOS locations to those obtained using Fastloc GPS from the same electronic tags on five species of pinnipeds: 9 California sea lions (Zalophus californianus), 4 Galapagos sea lions (Zalophus wollebaeki), 6 Cape fur seals (Arctocephalus pusillus pusillus), 3 Australian fur seals (A. p. doriferus) and 5 northern elephant seals (Mirounga angustirostris). These species encompass a range of marine habitats (highly pelagic vs coastal), diving behaviors (mean dive durations 2–21 min) and range of latitudes (equator to temperate). A total of 7,318 ARGOS positions and 27,046 GPS positions were collected. Of these, 1,105 ARGOS positions were obtained within five minutes of a GPS position and were used for comparison. The 68^th percentile ARGOS location errors as measured in this study were LC-3
0.49 km, LC-2 1.01 km, LC-1 1.20 km, LC-0 4.18 km, LC-A 6.19 km, LC-B 10.28 km.
Conclusions/Significance: The ARGOS errors measured here are greater than those provided by ARGOS, but within the range of other studies. The error was non-normally distributed with each LC highly right-skewed. Locations of species that make short duration dives and spend extended periods on the surface (sea lions and fur seals) had less error than species like elephant seals that spend more time underwater and have shorter surface intervals. Supplemental data (S1) are provided allowing the creation of density distributions that can be used in a variety of filtering algorithms to improve the quality of ARGOS tracking data.

Habitat suitability for marine fishes using presence-only modelling and multibeam sonar

Improved access to multibeam sonar and underwater video technology is enabling scientists to use spatially-explicit, predictive modelling to improve our understanding of marine ecosystems. With the growing number of modelling approaches available, knowledge of the relative performance of different models in the marine environment is required. Habitat suitability of 5 demersal fish taxa in Discovery Bay, south-east Australia, were modelled using 10 presence-only algorithms: BIOCLIM, DOMAIN, ENFA (distance geometric mean [GM], distance harmonic mean [HM], median [M], area-adjusted median [Ma], median + extremum [Me], area-adjusted median + extremum [Mae] and minimum distance [Min]), and MAXENT. Model performance was assessed using kappa and area under curve (AUC) of the receiver operator characteristic. The influence of spatial range (area of occupancy) and environmental niches (marginality and tolerance) on modelling performance were also tested. MAXENT generally performed best, followed by ENFA-GM and -HM, DOMAIN, BIOCLIM, ENFA-M, -Min, -Ma, -Mae and -Me algorithms. Fish with clearly definable niches (i.e. high marginality) were most accurately modelled. Generally, Euclidean distance to nearest reef, HSI-b (backscatter), rugosity and maximum curvature were the most important variables in determining suitable habitat for the 5 demersal fish taxa investigated. This comparative study encourages ongoing use of presence-only approaches, particularly MAXENT, in modelling suitable habitat for demersal marine fishes.

Restoration of the habitat-forming fucoid alga Hormosira banksii at effluent-affected sites : competitive exclusion by coralline turfs

Fucoid algae often dominate intertidal rocky shores, providing habitat and modifying ecosystem resources for other species, but are susceptible to discharge of sewage effluent. In this study we assessed the potential for competition from coralline turfs to inhibit restoration of the intertidal fucoid macroalga Hormosira banksii at sites associated with an ocean outfall a scenario of improving following water quality in the nearshore coastal environment. The percentage cover and number of individuals of H. banksii were negatively correlated with both the percentage cover and turf height of Corallina officinalis. In contrast, H. banksii was positively associated with rocky substrata and recruited well to rock-surface substrata. Importantly, there appears to be a threshold abundance where the percentage cover of H. banksii rarely reaches above 20% cover amongst coralline turfs with >40% cover. These data support a model of alternative community states: H. banksii dominated canopy on rocky substrata versus C. officinalis turf. In field and laboratory experiments, extensive coralline turfs (up to 4 cm thick) were shown to inhibit recruitment of H. banksii. This study shows competitive exclusion by coralline turfs may limit the successful restoration of habitat provided by H. banksii to shores that have been affected by sustained discharge of secondarily treated sewage effluent. We suggest potential strategies for management authorities to consider when seeking ways of restoring fucoid communities affected by anthropogenic disturbances such as wastewater disposal.

Habitat fragmentation disrupts the demography of a widespread native mammal

Fragmentation theory predicts that population persistence should be positively correlated with the size of habitat fragments. The patterns of occurrence of many species are consistent with this prediction, but the demographic processes that determine how species respond to fragmentation are poorly understood. In addition, habitat quality may interact with fragment size as an influence on demographic performance. We investigated these predictions for the native bush rat Rattus fuscipes by testing the following hypotheses: 1) population performance (i.e. viability as determined by various demographic parameters) is positively correlated with fragment size; and 2) population performance is positively correlated with habitat quality. Populations of R. fuscipes were censused in two large (>49 ha) and eight small (<2.5 ha) forest fragments in an agricultural region of southeastern Australia. Fragments with high and low quality habitat were included in each size category. Fragment size influenced multiple aspects of population demography; populations in large fragments had higher densities, older age structures, received more potential immigrants, and were more likely to recruit adults than those in small fragments. Reproductive patterns were more predictable in large fragments. Habitat quality per se had less marked effects; adult females were heavier and subadults more prevalent in fragments with high quality habitat. However, high quality habitat enhanced population performance in small fragments more so than in large ones. Despite being widespread in the study area, R. fuscipes populations are profoundly impacted by habitat fragmentation, with population performance declining with fragment size. Studies based on patterns of species occurrence should be interpreted with caution as they may mask critical processes occurring at the population level. For a thorough understanding of the effects of habitat fragmentation, population-level studies are required.

Remotely sensed hydroacoustics and observation data for predicting fish habitat suitability

This paper investigates the use of using remotely sensed observation and full coverage hydroacoustic datasets to quantify habitat suitability for a marine demersal fish, the blue-throated wrasse. Because of issues surrounding the detection of species using remotely sensed video techniques, the application of presence-only techniques are well suited for modeling demersal fish habitat suitability. Ecological-Niche Factor Analysis is used to compare analyses conducted using seafloor variables derived from hydroacoustics at three spatial scales; fine (56.25 m²), medium (506.25 m²) and coarse (2756.25 m²), to determine which spatial scale was most influential in predicting blue-throated wrasse habitat suitability. The coarse scale model was found to have the best predictive capabilities with a Boyce Index of 0.80±0.26. The global marginality and specialization values indicated that, irrespective of spatial scale, blue-throated wrasse prefer seafloor characteristics that are different to the mean available within the study site, but exhibit a relatively wide niche. Although variable importance varied over the three spatial scale models, blue-throated wrasse showed a strong preference for regions of shallow water, close to reef, with high rugosity and maximum curvature and low HSI-B values. Generally the spatial patterns in habitat suitability were well represented in the Marine National Park compared to adjacent waters. However, some significant differences in spatial patterns were observed. Interspersion and Juxtaposition Indexes for unsuitable and highly suitable habitat were significantly smaller inside the Marine National Park, while the Mean Shape Index of unsuitable habitat in the Marine National Park was significantly larger.

Habitat fragmentation and landscape change

This chapter begins by summarizing the conceptual approaches used to understand conservation in fragmented landscapes. We then examine the biophysical aspects of landscape change, and how such change affects species and communities, posing two main questions: (i) what are the implications for the patterns of occurrence of species and communities?; and (ii) how does landscape change affect processes that influence the distribution and viability of species and communities. The chapter concludes by identifying the kinds of actions that will enhance the conservation of biota in fragmented landscapes.

Spamcooling : a parallel heterogeneous ensemble spam filtering system based on active learning techniques

Anti-spam technology is developing rapidly in recent years. With the emerging applications of machine learning in diverse fields, researchers as well as manufacturers around the world have attempted a large number of related algorithms to prevent spam. In this paper, we designed an effective anti-spam protection system, SpamCooling, based on the mechanism of active learning and parallel heterogeneous ensemble learning techniques. The system adopts a batch method to filter spam and can be easily incorporated with existing mail clients (MUA). It can actively obtain user feedbacks for providing users with personalized spam filtering experiences. The parallel heterogeneous ensemble method can help system achieve high spam detection rate as well as low ham misclassification rate.

Estimating survival and movements using both live and dead recoveries : a case study of oystercatchers confronted with habitat change

1. Animals facing partial habitat loss can try to survive in the remaining habitat or emigrate. Effects on survival and movements should be studied simultaneously since survival rates may be underestimated if emigrants are not considered, and since emigrants may experience reduced survival.

2. We analysed movements and survival of adult wintering oystercatcher Haematopus ostralegus in response to the 1986–1987 partial closure of the Oosterschelde in the Dutch Delta. This reduced by one-third the tidal area of this major European wintering area for waders.

3. We developed a novel variant of a multistate capture–recapture model allowing simultaneous estimation of survival and movement between sites using a mixture of data (live recaptures and dead recoveries). We used a two-step process, first estimating movements between sites followed by site-specific survival rates.

4. Most birds were faithful to their ringing site. Winter survival was negatively affected by winter severity and was lowest among birds changing wintering site (i.e. moving outside of the Oosterschelde).

5. During mild winters, survival rates were very high, and similar to before the closure in both changed and unchanged sectors of the Oosterschelde. However, the combined effect of habitat loss with severe winters decreased the survival of birds from changed sectors and induced emigration.

6. The coastal engineering project coincided with three severe winters and high food stock, making assessment of its effects difficult. However, the habitat loss seems to have had less impact on adult survival and movements than did winter severity.

7. Synthesis and applications. Human-induced habitat change may result in population decline through costly emigration or reduced survival or reproduction of individuals that stay. Long-term monitoring of marked individuals helps to understand how populations respond to environmental change, but site-specific survival and movement rates should be integrated in the same model in order to maximize the information yield. Our modelling approach facilitates this because it allows the inclusion of recoveries from outside the study area.

The habitat value of Gorse Ulex europaeus L. and Hawthorn Crataegus monogyna Jacq. for birds in Quarry Hills Bushland Park, Victoria

Weeds are one of the primary threats to biodiversity; however, their impacts on wildlife can vary. This research investigated the habitat value of Gorse Ulex europaeus L. and Hawthorn Crataegus monogyna Jacq. and the impacts of its removal on birds in a bushland park in Victoria. The area search method was used to survey birds in vegetation dominated by these two weeds, in native vegetation and in areas where a weed removal program was undertaken; this included revegetated areas. The highest bird species richness and abundance was found in sites dominated by the weeds. At sites where the weed removal program was in the early stages, a much lower species richness and abundance occurred. The final stage of the weed removal program, where revegetated areas were older than five years, supported high richness and abundance of birds, but not as high as that of sites dominated by the weeds; nor was the composition the same. Thus, even after five years, revegetation may not provide for the bird community that was originally supported by weeds. This is an important weed management consideration in this park, and should be for weed removal projects elsewhere