Supervised Learning Techniques

'El Diario de Juárez' is a local newspaper in a city of 1.5 million Spanish-speaking inhabitants that publishes texts of which citizens read them on both a website and an RSS (Really Simple Syndication) service. This research applies natural-language-processing and machine-learning algorithms to the news provided by the RSS service in order to classify them based on whether they are about a traffic incident or not, with the final intention of notifying citizens where such accidents occur. The classification process explores the bag-of-words technique with five learners (Classification and Regression Tree (CART), Naïve Bayes, kNN, Random Forest, and Support Vector Machine (SVM)) on a class-imbalanced benchmark; this challenging issue is dealt with via five sampling algorithms: synthetic minority oversampling technique (SMOTE), borderline SMOTE, adaptive synthetic sampling, random oversampling, and random undersampling. Consequently, our final classifier reaches a sensitivity of 0.86 and an area under the precision-recall curve of 0.86, which is an acceptable performance when considering the complexity of analyzing unstructured texts in Spanish.

The increasing popularity of smart wearable devices can be attributed to progress in research towards human activity recognition. With the use of tiny sensing units, the different human activities of day-today life can be identified with a high accuracy. However, as a fitness tracking product, smart wearables are not always accurate in determining actual physical motion details. For example, recognizing walk activity may not always be possible when the monitoring device is held in hand, or kept in pockets. For precise recognition of walk activity, movement of legs needs to be monitored. However, the movement of legs while walking must be distinguished from simple leg swing activities. The present work designs an IMU based sensing system that can prevent false identification of a mimicked walk or leg swing in sitting posture as a real walk activity, using conventional and convolutional deep learning algorithms. The system shows remarkable capability of identifying an actual walk from a mimicked walk activity using CNN 95% of the time.

The current body of literature lacks studies related to organizational managers' classification of systems thinking (ST) skills based on both their overall systemic tendency and the organizational ownership structure. The purpose of this study is to assess and classify the ST skills of senior managers who currently work in a complex business environment. Initially, we clustered managers' overall systemic thinking (OST) using the Bayesian latent class analysis (BLCA) method into two distinct clusters: managers with upper OST (holistic thinker) and managers with lower OST (reductionist thinker). Further, we classified managers' ST skills into two predefined classes to understand the characteristics of each group better. A total of 51 senior managers from two different organizational structures participated in this study. Results show that the ST skills of managers in public are more towards the upper OST/holistic cluster, whereas managers from the private sector have an inclination towards the lower OST/reductionist cluster.

Stock worth measures have reliably pulled in the thought of various agents and authorities. Standard speculation holds that stock trades are sporadic in nature, and it is senseless to endeavour to envision them. Since various components are incorporated, anticipating the stock worth itself is a troublesome issue. Until further notice, the market continues like a popularity based machine, yet as time goes on, it acts like a measuring machine, so it can foresee exhibit designs over a progressively broadened time period. The utilization of machine learning advancement and various counts in stock worth examination and figure is a promising field. In this paper, we first rapidly plot the logical classification of the stock markets and taxonomy of stock market prediction methods. By then, we will focus on some investigation achieves stock assessment and deciding. We look at the particular, technical, fundamental, and long short term memory network strategies used in stock examination. Finally, we present a couple of challenges and investigation opportunities in this field.

Logistic regression analysis of factors contributing to bank deposit subscription and direct marketing success.

View project at:

https://github.com/JackKRasmus-Vorrath/Logistic_Regression_Direct-Marketing_Bank-Term-Deposit-Subscription

Machine Learning and Applications: An International Journal (MLAIJ) is a quarterly open access peer-reviewed journal that publishes articles which contribute new results in all areas of the machine learning. The journal is devoted to the publication of high quality papers on theoretical and practical aspects of machine learning and applications.The goal of this journal is to bring together researchers and practitioners from academia and industry to focus on machine learning advancements, and establishing new collaborations in these areas. Original research papers, state-of-the-art reviews are invited for publication in all areas of machine learning.
Authors are solicited to contribute to the journal by submitting articles that illustrate research results, projects, surveying works and industrial experiences that describe significant advances in the areas of machine learning.

Thanks to more powerful hardware and a new generation of learning algorithms, artificial intelligence is supporting the automation of a number of tasks and activities that are changing the job landscape as much as they have impacted on our everyday life. The first part of the chapter introduces artificial intelligence from its origins: its definition and its main research and application areas. The nature and the importance of machine learning for artificial intelligence applications are presented in the second part of the chapter. Existing approaches to machine learning are also classified and illustrated. The third part describes artificial intelligence tools and solutions by supported functionalities and automated tasks. Cases of applications in tourism are provided, from the best known and widely adopted, e.g., personal assistants, to the most challenging, i.e., semantic systems. Future trends and risks related to the applications of artificial intelligence are considered in the last part of the chapter.

— The rampant use of dangerous levels of formalin in food items has incited anxiety and concern amongst the Bangladeshi citizens. Sporadic monitoring, amalgamated with the paucity of well-grounded and inexpensive testing devices has created a need for the citizens to address the issue themselves. Impelled by promising results from a survey, FoodAlytics, a portable, economical and non-contact formalin detection system has been developed using locally sourced COTS components. Integrated with our own android application and bundled with the Grove VOC HCHO gas sensor, a recent innovation, our system has been able to detect up to 50 ppm of formalin. Polynomial regression, linear regression and the Levenberg-Marquardt algorithm, three supervised machine learning algorithms have been incorporated in our system to accurately predict the correct concentration of formalin at all temperatures. By applying logistic regression and support vector machines, our system is also able to correctly classify between artificially added and naturally formed formalin. The widespread adulteration of fresh fruits, vegetables and fish with harmful concentrations of formalin has become a severe issue in Bangladesh due to the health hazards associated with it [1][2]. Excessive concentrations of formalin has been found in almost all fruits available in the local market, thanks to lack of sufficient monitoring and raids [3]. Analogous to this issue, the authors decided to conduct a survey on 150 random subjects in and around Dhaka city to find out what our citizens think should be the impeccable solution to this problem. Figure 1 shows partial results of the survey. From Figure 1, it is clear that most of our citizens regularly shop for fruits and vegetables while more than half of them source their buys from roadside hawkers and local stalls/vans which are not monitored at all by our mobile courts. 66% of our subjects have confirmed that they think their buys are adulterated with toxic levels of formalin, while 10% of these subjects have checked for presence of formalin themselves and all of them found formalin in their fruits and vegetables. The outcome of the survey was further intensified when our subjects stated that the core problem was the lack of a dedicated formalin detection system that can be used by the general mass for detection of adulterated food items and reporting to concerned authorities. Furthermore, 91% of our subjects wanted a device in the market that the citizens themselves can use to check for adulterated food. Figure 2 outlines the specifications of the package as suggested by our subjects. Pertaining to the results of the survey, the authors have developed a low-cost, handy and comprehensible formalin detection package entitled " FoodAlytics " bundled with an android application. Supervised learning algorithms have been incorporated into our system to correctly classify between artificially added and naturally formed formalin in food items, as well as accurately infer the correct concentration of formalin present regardless of surrounding temperature. II. LITERATURE REVIEW Current handheld formalin detectors (Z-300) used by mobile courts in Bangladesh have a narrow concentration range (~1-30 ppm), practically importable, abstruse for everyday use, provides frequent erroneous readings and costs 1175 USD [4] [5]. In contrast, a chemical kit manufactured by Bangladesh The authors are grateful towards IEEE Bangladesh Section Humanitarian Activities Committee (IEEE BDS HAC) for providing necessary guidance and financially assisting in the making of this paper. Figure 1. Partial results of the survey (quoted in percentages). The inner chart shows the sources of fruits and vegetables for the subjects while the outer chart shows the frequency of shopping of fresh fruits and vegetables. Chart A shows that 61.1% of our subjects shop regularly (more than 2-4 times a week) while Chart B shows that 59% of our subjects shop from sources with no monitoring for adulteration. Figure 2. Specifications of the formalin detection system as suggested by our subjects. The figures are quoted in percentages.

The ‘special sauce’ that makes machine learning so interesting and what captures our imagination is not the ability of the machine to behave according to what you have already taught it in terms of explicit inputs, it is the implicit ability of the machine to construct relationships and patterns between input/output

Shill Bidding (SB) has been recognized as the predominant online auction fraud and also the most difficult to detect due to its similarity to normal bidding behavior. Previously, we produced a high-quality SB dataset based on actual auctions and effectively labeled the instances into normal or suspicious. To overcome the serious problem of imbalanced SB datasets, in this study, we investigate over-and under-sampling techniques through several instance-based classification algorithms. Thousands of auctions occur in eBay every day, and auction data may be sent continuously to the optimal fraud classifier to detect potential SB activities. Consequently , instance-based classification is appropriate for our particular fraud detection problem. According to the experimental results, incremental classification returns high performance for both over-and under-sampled SB datasets. Still, over-sampling slightly outperforms under-sampling for both normal and suspicious classes across all the classifiers.

Scope of the book:
This book focusses on the technical concepts of deep learning and its associated branch Neural Networks for the various dimensions of image processing applications. The proposed volume intends to bring together researchers to report the latest results or progress in the development of the above-mentioned areas. Since there is a deficit of books on this specific subject matter, the editors aim to provide a common platform for researchers working in this area to exhibit their novel findings.
Topics of Interest:
This book solicits contributions, which include the fundamentals in the field of Deep Artificial Neural Networks and Image Processing supported by case studies and practical examples. Each chapter is expected to be self-contained and to cover an in-depth analysis of real life applications of neural networks to image analysis.

The continually increasing number of documents produced each year necessitates ever improving information processing methods for searching, retrieving, and organizing text. Central to these information processing methods is document classification, which has become an important application for supervised learning. Recently the performance of these traditional classifiers has degraded as the number of documents has increased. This is because along with this growth in the number of documents has come an increase in the number of categories. This paper approaches this problem differently from current document classification methods that view the problem as multi-class classification. Instead we perform hierarchical classification using an approach we call Hierarchical Deep Learning for Text classification (HDLTex). HDLTex employs stacks of deep learning architectures to provide specialized understanding at each level of the document hierarchy.

In this work methods for performing time series prediction on complex real
world time series are examined. In particular series exhibiting non-linear or
chaotic behaviour are selected for analysis. A range of methodologies based
on Takens’ embedding theorem are considered and compared with more
conventional methods. A novel combination of methods for determining the
optimal embedding parameters are employed and tried out with multivariate
financial time series data and with a complex series derived from an
experiment in biotechnology. The results show that this combination of
techniques provide accurate results while improving dramatically the time
required to produce predictions and analyses, and eliminating a range of
parameters that had hitherto been fixed empirically. The architecture and
methodology of the prediction software developed is described along with
design decisions and their justification. Sensitivity analyses are employed to
justify the use of this combination of methods, and comparisons are made
with more conventional predictive techniques and trivial predictors showing
the superiority of the results generated by the work detailed in this thesis.

Supervised and unsupervised seismic facies classification methods are slowly gaining popularity in hydrocarbon exploration and production workflows. Unsupervised clustering is data driven, unbiased by the interpreter beyond the choice of input data and brings out the natural clusters present in the data. There are several competing unsupervised clustering techniques, each with advantages and disadvantages. In this dissertation, I demonstrate the use of various classification techniques on real 3D seismic data from various depositional environments. Initially, I use the popular unsupervised Kohonen self-organizing maps (SOMs) algorithms on case studies based on offshore deep-water Gulf of Mexico field and onshore heterogeneous Mississippian Chert reservoir from Oklahoma, USA. However, Kohonen SOM suffers from the absence of any proper convergence criterion and rules for parameter selection. These shortcomings are addressed by the more recent development of generative topographic mapping (GTM) algorithm. GTM is based on a probabilistic unsupervised classification technique. I extend the GTM workflow to consider multi-attribute inversion volumes and do seismic facies classification for a Barnett shale survey and another case study on Veracruz Basin survey in Southern Mexico that was acquired over a heterogeneous conglomerate reservoir.Finally, I introduce limited supervision using user defined clusters into both the SOM and GTM algorithms.
The final product of my dissertation is a suite of algorithms, workflows, user interfaces and user documentation allowing others to build upon and extend this research.

The key to the keys to immortality and eternal youth lies in the correct answer to the main question: How to naively discover new essential – but still hidden – features required for properly training novel adaptive supervised machine learning algorithms (SMLA)

Language acquisition - especially in human infants - is a problem that intrigues the layman and baffes the expert. This article takes a computational viewpoint toward the problem and investigates the problem of learnability of languages. We look at some results that show that learning is impossible even under very weak criterion. We next look at a framework in which learning takes place with a helpful teacher and demonstrate the role such a teacher can play in easing the learning problem. The discussion would define language learning formally, survey classical results and point toward recent advances in the field.

Sentiment analysis on Twitter offers possibilities of great interest to evaluate the currents of opinion disseminated through this medium. The huge volumes of texts require tools able to automatically process these messages without losing reliability. This paper describes two different types of approaching this problem. The first strategy is based on Supervised Learning processes, developed in the field of artificial intelligence. Its application requires some tools from natural language processing along with a classifed corpus as a starting point. The second approach is based on polarity dictionaries. SentiStrength tool is located in this line. It is increasingly applied to studies of Twitter in English. The paper assesses the most advanced studies using each of these approaches for analyzing tweets in Spanish. Finally, the advantages and limitations of each of these approaches for researching political communication are assessed. While supervised learning allows taking into account the context thanks to its ability to detect patterns of words, the researcher who uses this approach requires having data analyst skills to better refine the process. Instead, SentiStrength is more oriented to the semantic content of the terms of the message. It requires more of a competence in linguistics by the researcher. The main conclusion of this study is that both automated methods of analysis can not do without a demanding manual coding if they are to be used reliably in research.

Classification techniques classify the remotely sensed image by using reflectance properties of pixels. This paper presents a new approach to classify multispectral remotely sensed image. This approach classifies the multispectral image using frequencies of spectral bands' grey level values (DN values) in Histogram. It draws histogram for different spectral bands of the image. Then, it finds and separates the humps in histograms. This approach yields more meaningful classification for multi-modal or bi-modal histograms. It creates 3 potential centroids in each hump for each spectral band. More the number of humps, more would be potential centroids for classification. Different spectral bands have different peaks in their humps of histograms. It reads all the pixels of one peak of one band and draw the local histogram of other bands' grey level values using pixels read. This way, peak of one hump of one band can find corresponding peaks in local histogram and these peaks make a pixel that can be a potential centroid and some of these peak frequencies is the actual frequency of that centroid. Now, I choose extreme left and extreme right grey level values whose frequency is greater than or equal to the average frequency of that hump. As each hump of each spectral band has three grey level values, I can find three centroids for each hump of each spectral band. Duplicate centroids are eliminated from the list of centroids. The rest of the centroids are recursively iterated and centroids with lesser frequencies than the nearby centroids are eliminated. Later, algorithm uses gravitational force to find out two nearby centroids.

Not all instances in a data set are equally beneficial for inferring a model of the data. Some instances (such as outliers) are detrimental to inferring a model of the data. Several machine learning techniques treat instances in a data set differently during training such as curriculum learning, filtering, and boosting. However, an automated method for determining how beneficial an instance is for inferring a model of the data does not exist. In this paper, we present an automated method that orders the instances in a data set by complexity based on the their likelihood of being misclassified (instance hardness). The underlying assumption of this method is that instances with a high likelihood of being misclassified represent more complex concepts in a data set. Ordering the instances in a data set allows a learning algorithm to focus on the most beneficial instances and ignore the detrimental ones. We compare ordering the instances in a data set in curriculum learning, filtering and boosting. We find that ordering the instances significantly increases classification accuracy and that filtering has the largest impact on classification accuracy. On a set of 52 data sets, ordering the instances increases the average accuracy from 81% to 84%.

Machine learning has become popular today as so many of its algorithms are now commonly used in different data science projects in various industries especially in the health care sector. It is imperative for researchers and medical professionals to be assisted by machine learning methods in early detection of diseases such as heart disease which is one major killer of humans in our world today. Machine learning algorithms are excellent at learning from data, and since healthcare providers generate huge amount of data on a daily basis, these algorithms can thrive in this field. In this research study, a comparative analytical approach was taken in the determination of which algorithm performs better under the given condition. Various experiments were carried out using cross validation of 5 and 10 folds, to ensure that models created can generalize well enough. This study makes use of data from University of California, Irvine (UCI) machine learning database containing 303 instances with 14 attributes. The collected data is scaled using Min-Max normalization technique. Different popular models are built using supervised machine learning classification algorithms on the scaled data such as Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Logistic Regression (LR), Naïve Bayes (NB), Random Forest (RF), and Gradient Boosting ensemble method. These algorithms are also evaluated using standard performance metrics such as precision, recall, and F1-score. From the experiments carried out, it can be concluded that SVM performs better as it out performs the other algorithms.

During the past decade, the size of 3D seismic data volumes and the number of seismic attributes have increased
to the extent that it is difficult, if not impossible, for interpreters to examine every seismic line and time
slice. To address this problem, several seismic facies classification algorithms including k-means, self-organizing
maps, generative topographic mapping, support vector machines, Gaussian mixture models, and artificial neural
networks have been successfully used to extract features of geologic interest from multiple volumes. Although
well documented in the literature, the terminology and complexity of these algorithms may bewilder the average
seismic interpreter, and few papers have applied these competing methods to the same data volume. We have
reviewed six commonly used algorithms and applied them to a single 3D seismic data volume acquired over the
Canterbury Basin, offshore New Zealand, where one of the main objectives was to differentiate the architectural
elements of a turbidite system. Not surprisingly, the most important parameter in this analysis was the choice of
the correct input attributes, which in turn depended on careful pattern recognition by the interpreter. We found
that supervised learning methods provided accurate estimates of the desired seismic facies, whereas unsupervised
learning methods also highlighted features that might otherwise be overlooked.

Cloud computing has been widely adopted by application service providers (ASPs) and enterprises to reduce both capital expenditures (CAPEX) and operational expenditures (OPEX). Applications and services previously running on private data centers are now being migrated to private or public clouds. Since most of the ASPs and enterprises have globally distributed user bases, their services need to be distributed across multiple clouds, spread across the globe which can achieve better performance in terms of latency, scalability and load balancing. The shift has eventually led the research community to study multi-cloud environments. However, the widespread acceptance of such environments has been hampered by major security concerns. Firewalls and traditional rule-based security protection techniques are not sufficient to protect user-data in multi-cloud scenarios. Recently, advances in machine learning techniques have attracted the attention of the research community to build intrusion detection systems (IDS) that can detect anomalies in the network traffic. Most of the research works, however, do not differentiate among different types of attacks. This is, in fact, necessary for appropriate countermeasures and defense against attacks. In this paper, we investigate both detecting and categorizing anomalies rather than just detecting, which is a common trend in the contemporary research works. We have used a popular publicly available dataset to build and test learning models for both detection and categorization of different attacks. To be precise, we have used two supervised machine learning techniques, namely linear regression (LR) and random forest (RF). We show that even if detection is perfect, categorization can be less accurate due to similarities between attacks. Our results demonstrate more than 99% detection accuracy and categorization accuracy of 93.6%, with the inability to categorize some attacks. Further, we argue that such categorization can be applied to multi-cloud environments using the same machine learning techniques.

Motivation: The genotype assignment problem consists of predicting, from the genotype of an individual , which of a known set of populations it originated from. The problem arises in a variety of contexts, including wildlife forensics, invasive species detection and biodiversity monitoring. Existing approaches perform well under ideal conditions but are sensitive to a variety of common violations of the assumptions they rely on. Results: In this article, we introduce Mycorrhiza, a machine learning approach for the genotype assignment problem. Our algorithm makes use of phylogenetic networks to engineer features that encode the evolutionary relationships among samples. Those features are then used as input to a Random Forests classifier. The classification accuracy was assessed on multiple published empirical SNP, microsatellite or consensus sequence datasets with wide ranges of size, geographical distribution and population structure and on simulated datasets. It compared favorably against widely used assessment tests or mixture analysis methods such as STRUCTURE and Admixture, and against another machine-learning based approach using principal component analysis for dimensionality reduction. Mycorrhiza yields particularly significant gains on datasets with a large average fixation index (FST) or deviation from the Hardy-Weinberg equilibrium. Moreover, the phylogenetic network approach estimates mixture proportions with good accuracy. Availability and implementation: Mycorrhiza is released as an easy to use open-source python package at github.com/

A new Learning Vector Quantization classifier is proposed. The algorithm relies on a new training scheme for labeled sample vectors in feature space. Since weight or prototype vectors are conditioned to a well-known sliding-mode approach with use of a cost function to be minimized in terms of weight updates, new algorithm is called Optimized Generalized Learning Vector Quantization (OGLVQ). Consequently, weights are then associated to the proximity measure employed by conventional Generalized Learning Vector Quantization. New algorithm and some well-known predecessors are designed and tested for comparison with synthetic and publicly available datasets. From the experimental results, it is observed that the new classifier achieves faster training and is more successful and robust in generalizing labeled test samples picked from datasets studied than the counterparts it is compared to.

The climate change has caused threats to agricultural production; the extremes of temperature and humidity, and other abiotic stresses are contributing factors to the etiology of disease and pest on crops. About the matter, recent research efforts have focused on predicting disease and pest crops using computer science techniques such as supervised learning algorithms. Therefore in this paper, we present an overview of supervised learning algorithms commonly used in agriculture for the detection of pests and diseases in crops such as corn, rice, coffee, mango, peanut, and tomato, among others, with the aim of selecting the algorithms that give the best performance
for the agricultural sector.

This project aims at developing, validating, and testing several classification statistical models that could predict whether or not an office room is occupied using several data features, namely temperature (◦C), light (lx), humidity (%), CO2 (ppm), and a humidity ratio. The data is modeled using classification techniques i.e. Logistic regression, Classification tree, Bagging-Random forest, and Gradient boosted trees.
These models were trained and then after evaluated against validation and test sets and using confusion matrices to obtain classification and misclassification rates. The logistic model was trained using glmnet R package, Tree package for classification tree model, randomForest for both Bagging and Random Forest Models, and gbm package for Gradient Boosted Model.
The best accuracy was obtained from the Random Forest Model with a classification rate of 93.21% when it was evaluated against the test set. Light sensor is also the most significant variable in predicting whether the office room is occupied or not, this was observed in all the five models.

We propose an evolutionary method for optimising both the architecture and the synaptic weights of single hidden-layer feed forward neural networks. Based on evolutionary strategies, this method uses new genetic operators of mutation and recombination in order to evolve a population of candidate solutions in the form of neural networks with different architectures. Experimental results are presented to demonstrate the effectiveness of the proposed method in both classification and prediction problems. These results concern six well-known benchmark problems and are compared to those produced for the same problems by three other methods.

Coronavirus adalah kumpulan virus yang bisa menginfeksi sistem pernapasan. Pada banyak kasus, virus ini hanya menyebabkan infeksi pernapasan ringan, seperti flu. Namun, virus ini juga bisa menyebabkan infeksi pernapasan berat, seperti infeksiparu-paru (pneumonia). Banyak sistem kesehatan di dunia telah runtuh karena kapasitas yang terbatas dan peningkatan dramatis dari pasien yang terjangkit kasus COVID-19. Yang muncul adalah perlunya menemukan metode yang efisien, cepat dan akurat untuk mengurangi kelebihan upaya ahli radiologi untuk mendiagnosis kasus yang diduga. Rancangan ini menyajikan kombinasi metode klasifikasi supervised learning untuk membedakan antara coronavirus COVID-19, pneumonia dan paru-paru yang sehat.

The main aim in network anomaly detection is effectively spotting hostile events within the traffic pattern associated to network operations, by distinguishing them from normal activities. This can be only accomplished by acquiring the a-priori knowledge about any kind of hostile behavior that can potentially affect the network (that is quite impossible for practical reasons) or, more easily, by building a model that is general enough to describe the normal network behavior and detect the violations from it. Earlier detection frameworks were only able to distinguish already known phenomena within traffic data by using pre-trained models based on matching specific events on pre-classified chains of traffic patterns. Alternatively, more recent statistics-based approaches were able to detect outliers respect to a statistic idealization of normal network behavior. Clearly, while the former approach is not able to detect previously unknown phenomena (zero-day attacks) the latter one has limited effectiveness since it cannot be aware of anomalous behaviors that do not generate significant changes in traffic volumes. On the other hand, machine learning allows the development of adaptive, non-parametric detection strategies that are based on " understanding " the network dynamics by acquiring through a proper training phase a more precise knowledge about normal or anomalous phenomena in order to classify and handle in a more effective way any kind of behavior that can be observed on the network. Accordingly, we present a new anomaly detection strategy based on supervised machine learning, and more precisely on a batch relevance-based fuzzyfied learning algorithm known as U-BRAIN, aiming at understanding through inductive inference the specific laws and rules governing normal or abnormal network traffic, in order to reliably model its operating dynamics. This proposal appears to be promising both in terms of identification accuracy and robustness/flexibility when coping with uncertainty in the detection/classification process, as verified through extensive evaluation experiments.

We propose a method which, given a document to be classified, automatically generates an ordered set of appropriate descriptors extracted from a thesaurus. The method creates a Bayesian network to model the thesaurus and uses probabilistic inference to select the set of descriptors having high posterior probability of being relevant given the available evidence (the document to be classified). Our model can be used without having preclassified training documents, although it improves its performance as long as more training data become available. We have tested the classification model using a document dataset containing parliamentary resolutions from the regional Parliament of Andalucía at Spain, which were manually indexed from the Eurovoc thesaurus, also carrying out an experimental comparison with other standard text classifiers.

Data plays a key role in the design of expert and intelligent systems and therefore, data preprocessing appears to be a critical step to produce high-quality data and build accurate machine learning models. Over the past decades, increasing attention has been paid towards the issue of class imbalance and this is now a research hotspot in a variety of fields. Although the resampling methods, either by under-sampling the majority class or by over-sampling the minority class, stand among the most powerful techniques to face this problem, their strengths and weaknesses have typically been discussed based only on the class imbalance ratio. However, several questions remain open and need further exploration. For instance, the subtle differences in performance between the over-and under-sampling algorithms are still undercomprehended, and we hypothesize that they could be better explained by analyzing the inner structure of the data sets. Consequently, this paper attempts to investigate and illustrate the effects of the resampling methods on the inner structure of a data set by exploiting local neighborhood information, identifying the sample types in both classes and analyzing their distribution in each resampled set. Experimental results indicate that the resampling methods that produce the highest proportion of safe samples and the lowest proportion of unsafe samples correspond to those with the highest overall performance. The significance of this paper lies in the fact that our findings may contribute to gain a better understanding of how these techniques perform on class-imbalanced data and why over-sampling has been reported to be usually more efficient than under-sampling. The outcomes in this study may have impact on both research and practice in the design of expert and intelligent systems since a priori knowledge about the internal structure of the imbalanced data sets could be incorporated to the learning algorithms.

Email has become an important means of electronic communication but the viability of its usage is marred by Un-solicited Bulk Email (UBE) messages. UBE poses technical and socio-economic challenges to usage of emails. Besides, the definition and understanding of UBE differs from one person to another. To meet these challenges and combat this menace, we need to understand UBE. Towards this end, this paper proposes a classifier for UBE documents. Technically, this is an application of un-structured document classification using text content analysis and we approach it using supervised machine learning technique. Our experiments show the success rate of proposed classifier is 98.50%. This is the first formal attempt to provide a novel tool for UBE classification and the empirical results show that the tool is strong enough to be implemented in real world.