Sequential Minimal Optimization (SMO)

Penelitian ini akan berfokus dalam membandingkan kinerja tiga algoritma klasifikasi data mining dalam mengklasifikasi pelanggan baru ke dalam klaster berdasarkan perilaku. Algoritma atau metodologi yang digunakan adalah Support Vector Machine (SVM), Decision Tree, dan Naïve Bayes. Penelitian sebelumnya telah berhasil mengimplementasikan algoritma K-Means Clustering untuk mengelompokkan data pelanggan dan menemukan kelompok-kelompok spesifik dengan karakteristik yang memiliki kemiripan antara satu dan lainnya. Penelitian lanjutan perlu dilakukan untuk mengeksplor lebih jauh mengenai data mining dan membandingkan kinerja algoritma dan pengujian model ke dalam sistem e-commerce nyata. Tahapan yang akan dilakukan pada penelitian ini akan meliputi pengumpulan data, prapemrosesan data, pembagian data untuk pelatihan dan pengujian, lalu membandingkan tingkat akurasi yang dihasilkan algoritma Support Vector Machine (SVM), Decision Tree, dan Naïve Bayes. Setelah diketahui algoritma yang paling unggul, akan diimplementasikan ke dalam aplikasi e-commerce.

Sequential minimal optimization (SMO) algorithm is one of the simplest decomposition methods for learning of support vector machines (SVMs). Keerthi and Gilbert have recently studied the convergence property of SMO algorithm and given a proof that SMO algorithm always stops within a finite number of iterations. In this paper, we point out the incompleteness of their proof and give a more rigorous proof.

Sequential minimal optimization (SMO) algorithm is one of the simplest decomposition methods for learning of support vector machines (SVMs). Keerthi and Gilbert have recently studied the convergence property of SMO algorithm and given a proof that SMO algorithm always stops within a finite number of iterations. In this paper, we point out the incompleteness of their proof and give a more rigorous proof.

Malignant melanoma represents the most dangerous type of skin cancer. In this study we present an ensemble classification scheme, employing the mutual information, the cross-correlation and the clustering based on proximity of image features methods, for early stage assessment of melanomas on plain photography images. The proposed scheme performs two main operations. First, it retrieves the most similar, to the unknown case, image samples from an available image database with verified benign moles and malignant melanoma cases. Second, it provides an automated estimation regarding the nature of the unknown image sample based on the majority of the most similar images retrieved from the available database. Clinical material comprised 75 melanoma and 75 benign plain photography images collected from publicly available dermatological atlases. Results showed that the ensemble scheme outperformed all other methods tested in terms of accuracy with 94.9±1.5%, following an external cross-val...

Malignant melanoma represents the most dangerous type of skin cancer. In this study we present an ensemble classification scheme, employing the mutual information, the cross-correlation and the clustering based on proximity of image features methods, for early stage assessment of melanomas on plain photography images. The proposed scheme performs two main operations. First, it retrieves the most similar, to the unknown case, image samples from an available image database with verified benign moles and malignant melanoma cases. Second, it provides an automated estimation regarding the nature of the unknown image sample based on the majority of the most similar images retrieved from the available database. Clinical material comprised 75 melanoma and 75 benign plain photography images collected from publicly available dermatological atlases. Results showed that the ensemble scheme outperformed all other methods tested in terms of accuracy with 94.9±1.5%, following an external cross-val...

As most of human activities are being moved to cyberspace, phishers and other cybercriminals are making the cyberspace unsafe by causing serious risks to users and businesses as well as threatening global security and economy. Nowadays, phishers are constantly evolving new methods for luring user to reveal their sensitive information. To avoid falling victim to cybercriminals, a phishing detection algorithms is very necessary to be developed. Machine learning or data mining algorithms are used for phishing detection such as classification that categorized cyber users in to either malicious or safe users or regression that predicts the chance of being attacked by some cybercriminals in a given period of time. Many techniques have been proposed in the past for phishing detection but due to dynamic nature of some of the many phishing strategies employed by the cybercriminals, the quest for better solution is still on. In this paper, we propose a new phishing detection model based on Ex...

Cybercriminals exploit malicious URLs as a distribution channel to spread harmful software across the internet. They take advantage of vulnerabilities in browsers to install malicious software with the aim of gaining remote access to the victims' computers. Typically, this malicious software aims to gain access to networks, steal sensitive information, and silently monitor targeted computer systems. In this research, a data mining approach known as Classification Based on Association (CBA) is employed to detect malicious URLs using both the URL itself and the features of the presented web pages. The CBA algorithm utilizes a training dataset of URLs as historical data to discover association rules that can be used to create an accurate classifier. By detecting dangerous URLs and malicious software, this contribution can assist organizations and individual users in enhancing the security of their computer systems and networks, thereby protecting sensitive data and reducing the ris...

Support Vector Machines are a powerful machine learning technology, but the training process involves a dense quadratic optimization problem and is computationally challenging. A parallel implementation of Support Vector Machine training has been developed, using a combination of MPI and OpenMP. Using an interior point method for the optimization and a reformulation that avoids the dense Hessian matrix, the structure of the augmented system matrix is exploited to partition data and computations amongst parallel processors efficiently. The new implementation has been applied to solve problems from the PASCAL Challenge on Large Scale Learning. We show that our approach is competitive, and is able to solve problems in the Challenge many times faster than other parallel approaches. We also demonstrate that the hybrid version performs more efficiently than the version using pure MPI.

The k-fold cross-validation is commonly used to evaluate the effectiveness of SVMs with the selected hyper-parameters. It is known that the SVM k-fold cross-validation is expensive, since it requires training k SVMs. However, little work has explored reusing the h th SVM for training the (h + 1) th SVM for improving the efficiency of k-fold cross-validation. In this paper, we propose three algorithms that reuse the h th SVM for improving the efficiency of training the (h + 1) th SVM. Our key idea is to efficiently identify the support vectors and to accurately estimate their associated weights (also called alpha values) of the next SVM by using the previous SVM. Our experimental results show that our algorithms are several times faster than the k-fold cross-validation which does not make use of the previously trained SVM. Moreover, our algorithms produce the same results (hence same accuracy) as the k-fold cross-validation which does not make use of the previously trained SVM.

Second order SMO represents the state-of-the-art in SVM training for moderate size problems. In it, the solution is attained by solving a series of subproblems which are optimized w.r.t just a pair of multipliers. In this paper we will illustrate how SMO works in a two stage fashion, setting first the values of the bounded multipliers to the penalty factor C and proceeding then to adjust the non-bounded multipliers. Furthermore, during this second stage the selected pairs for update often appear repeatedly during the algorithm. Taking advantage of this, we shall propose a procedure to combine previously used descent directions that results in much fewer iterations in this second stage and that may also lead to noticeable savings in kernel operations.

In this brief, we give a new proof of the asymptotic convergence of the sequential minimum optimization (SMO) algorithm for both the most violating pair and second order rules to select the pair of coefficients to be updated. The proof is more self-contained, shorter, and simpler than previous ones and has a different flavor, partially building upon Gilbert's original convergence proof of its algorithm to solve the minimum norm problem for convex hulls. It is valid for both support vector classification (SVC) and support vector regression, which are formulated under a general problem that encompasses them. Moreover, this general problem can be further extended to also cover other support vector machines (SVM)-related problems such as -SVC or one-class SVMs, while the convergence proof of the slight variant of SMO needed for them remains basically unchanged.

Support vector regression (SVR) is a powerful tool in modeling and prediction tasks with widespread application in many areas. The most representative algorithms to train SVR models are Shevade et al.'s Modification 2 and Lin's WSS1 and WSS2 methods in the LIBSVM library. Both are variants of standard SMO in which the updating pairs selected are those that most violate the Karush-Kuhn-Tucker optimality conditions, to which LIBSVM adds a heuristic to improve the decrease in the objective function. In this paper, and after presenting a simple derivation of the updating procedure based on a greedy maximization of the gain in the objective function, we show how cycle-breaking techniques that accelerate the convergence of support vector machines (SVM) in classification can also be applied under this framework, resulting in significantly improved training times for SVR.

In the recent years support vector machines (SVMs) have been successfully applied to solve a large number of classification problems. Training an SVM, usually posed as a quadratic programming (QP) problem, often becomes a challenging task for the large data sets due to the high memory requirements and slow convergence. We propose to apply boosting to Platt's Sequential Minimal Optimization (SMO) algorithm and to use resulting Boost-SMO method for speeding and scaling up the SVM training. Experiments on three commonly used benchmark data sets show that Boost-SMO achieves classification accuracy comparable to conventional SMO but is a factor of 3 to 10 faster. The speed-up could easily be orders of magnitude on the larger data sets.

The use of digital banks in Indonesia has rapidly increased in recent years in response to the adoption of new technologies and changes in consumer behavior. User responses to digital banks vary depending on their experience throughout their transactions on the application, which may result in satisfaction or dissatisfaction. Social media platforms such as Twitter have become a space for companies to obtain textual data related to customer reviews and their brand image. In this study, data obtained from Twitter have undergone the stages of data crawling and data cleaning. The subsequent stages involved classification using the Naïve Bayes algorithm and word cloud visualization to identify the most commonly used words based on user responses. The results of this study indicate that users' positive sentiment towards digital banks is influenced by the application's ease of use, while dissatisfaction is caused by technical constraints experienced during the administrative process. The positive, negative, and neutral sentiments in this study are used to identify business opportunities for digital banks and practical implications for future digital banking services.

Indonesia, as a predominantly Muslim country, experiences a high demand for halal products due to its Muslim consumer base. This research focuses on understanding the perceptions of Indonesian society towards businesses, particularly in the food and beverage sector, that lack halal certification. The specific company under investigation is Mixue Indonesia. The study utilizes sentiment analysis through the Naive Bayes method, aided by RapidMiner software, to analyze public comments on Twitter regarding Mixue Indonesia's halal certification issue. Out of the 200 comments analyzed, the sentiment analysis revealed that approximately 122 comments expressed negative sentiments towards Mixue's halal certification. The findings indicate a notable concern and dissatisfaction among the Indonesian community regarding Mixue Indonesia's halal certification status. To address this issue, it is recommended that Mixue Indonesia takes immediate steps to obtain halal certification for their products. Additionally, future research should explore the reasons behind the negative sentiments and further investigate the impact of halal certification on consumer behavior and purchasing decisions. Understanding these factors will enable companies like Mixue Indonesia to meet the demands and expectations of their Muslim consumer base effectively.

Indonesia, as a predominantly Muslim country, experiences a high demand for halal products due to its Muslim consumer base. This research focuses on understanding the perceptions of Indonesian society towards businesses, particularly in the food and beverage sector, that lack halal certification. The specific company under investigation is Mixue Indonesia. The study utilizes sentiment analysis through the Naive Bayes method, aided by RapidMiner software, to analyze public comments on Twitter regarding Mixue Indonesia's halal certification issue. Out of the 200 comments analyzed, the sentiment analysis revealed that approximately 122 comments expressed negative sentiments towards Mixue's halal certification. The findings indicate a notable concern and dissatisfaction among the Indonesian community regarding Mixue Indonesia's halal certification status. To address this issue, it is recommended that Mixue Indonesia takes immediate steps to obtain halal certification for their products. Additionally, future research should explore the reasons behind the negative sentiments and further investigate the impact of halal certification on consumer behavior and purchasing decisions. Understanding these factors will enable companies like Mixue Indonesia to meet the demands and expectations of their Muslim consumer base effectively.

Dunia pada saat ini tengah mengalami perkembangan pada bidang teknologi dan komunikasi secara massif, terlebih pada masa pandemi saat ini yang mengharuskan semua proses pembelajaran bahkan bekerja secara daring. Hal ini yang memicu banyaknya kejahatan di dunia internet. Salah satunya yaitu mencuri data pengguna internet melalui sebuah website palsu yang dibangun seperti asli atau disebut juga website phising. Pada penelitian ini, untuk mengatasi maraknya website phising di dunia maya, maka diperlukan suatu model klasifikasi untuk mendeteksi website yang terindikasi phising dengan menggunakan kinerja terbaik dari salah satu algoritma klasifikasi logistic regression dan random forest. Untuk meningkatkan kinerja algoritma klasifikasi dilakukan seleksi fitur menggunkan metode correlation-based feature selection (CFS) untuk menyeleksi atribut yang paling berpengaruh dalam mendeteksi web phising. Berdasarkan hasil uji coba, penerapan algoritma klasifikasi logistic regression dan random forest dalam klasifikasi web phising dihasilkan akurasi sebesar 93,035 % dan 96.834 %, setelah dilakukan seleksi fitur dengan CFS akurasi yang dihasilkan menjadi 92.718 % dan 97.015 %. Dari uji coba terjadi peningkatan akurasi pada random forest sebesar 0,181 % dan terjadi penurunan yang tidak signifikan pada logistic regression. Hasil uji coba membuktikan bahwa seleksi fitur dengan CFS dapat menghilangkan atribut redundan dan dihasilkan akurasi algoritma klasifikasi yang tidak jauh berbeda ketika atribut lengkap dan Random Forest memiliki akurasi lebih baik setelah menggunakan CFS .

Pemilihan fitur merupakan salah satu bagian penting dan teknik yang sering digunakan dalam praproses penggalian data yang membawa efek langsung untuk mempercepat algoritma penggalian data dan meningkatkan kinerja pertambangan seperti akurasi prediksi dan hasil yang komprehensif. Penelitian ini membahas mengenai pemilihan subset fitur dalam klasifikasi loyalitas pelanggan terhadap merek bagi pengguna fast moving consumer goods(dalam penelitian ini mengambil studi kasus pada salah satu produknya yaitu mie instan) dan melakukan analisis terhadap fitur-fitur yang mempengaruhi performa klasifikasi pohon keputusan. Data yang digunakan pada penelitian ini merupakan hasil penyebaran kuisoner kepada para pelanggan mie instan di Propinsi Lampung. Data yang diperoleh memiliki fitur yang bersifat heterogen, untuk itu dilakukan pengubahan fitur menjadi fitur homogen. Dalam penelitian ini, mengkombinasikan metode UFT (unsupervised feature transformation) dan metode DMI (dynamic mutual information) untuk seleksi fitur. Metode UFT digunakan untuk transformasi fitur non-numerik menjadi fitur numerik, sehingga fitur yang bersifat heterogen menjadi fitur homogen. Metode DMI digunakan untuk pemilihan fitur. Hasil transformasi fitur diklasifikasikan menggunakan algoritmapohon keputusan. Hasil klasifikasi digunakan untukmelakukan perbandingan performa antara dataset sebelum pemilihan fitur, setelah dilakukanpemilihan fitur menggunakan metode DMI, p-Value dan perkiraan peneliti. Dari hasil pengujian terhadap model prediksi klasifikasi diperoleh fiturfitur yang mempengaruhi performa klasifikasi pohon keputusanloyalitas pelanggan. Peningkatan performa tersebut dapat dilihat pada pengimplementasian metode pemilihan fitur DMIdengan jumlah fitur sebanyak lima. Nilai akurasi, presisi, recall dan f-measure mengalami peningkatan bila dibandingkan dengan penggunaan seluruh fitur (sebelum dilakukan pemilihan fitur), metode pemilihan fitur p-value dan hasil perkiraan, masing-masing nilai tersebut secara berturutturut adalah sebesar 76.68%, 74.4%, 76.7% dan 73.5%.Fitur-fitur yang berpengaruh tersebut antara lain jumlah pengeluaran, rata-rata konsumsi, usia, alamat dan alasan berpindah merek. Kata kunci: klasifikasi, loyalitas pelanggan, mutual informasi, transformasi fitur vi

Cancer is currently one of the main health issues in the world. Among different varieties of cancers, skin cancer is the most common cancer in the world and accounts for 75% of the world's cancer. Indeed, skin cancer involves abnormal changes in the outer layer of the skin. Although most people with skin cancer recover, it is one of the major concerns of people due to its high prevalence. Most types of skin cancers grow only locally and invade adjacent tissues, but some of them, especially melanoma (cancer of the pigment cells), which is the rarest type of skin cancer, may spread through the circulatory system or lymphatic system and reach the farthest points of the body. Many papers have been reviewed about the application of image processing in cancer detection. In this paper, the automatic skin cancer detection and also different steps of such a process have been discussed based on the implantation capabilities.

In the era of globalization, prediction of financial distress is of interest not only to managers but also to external stakeholders of a company. The stakeholders are continuously seeking the optimal solution for performance forecasting, as a way to rationalize the decision-making process. The recent past shows that financial stability of companies is at the stake. Stockholders, Managers, Creditor and employees of the business are always concerned about financial stability of the companies. The most frequently tool for financial analysis is financial ratios. However, financial ratios are no-longer proved appropriate for „Stockholders‟ equity position and creditors‟ claims. Stakeholder‟s have concerns about the consequences of financial distress for companies, and controls of capital adequacy through the regulatory capital requirement (Mingo, 2000). This shared interest creates persistent investigations and continuing attempts to answer an incessant question that how financial distre...

The nearest point problem (NPP), i.e., finding the closest points between two disjoint convex hulls, has two classical solutions, the Gilbert-Schlesinger-Kozinec (GSK) and Mitchell-Dem'yanov-Malozemov (MDM) algorithms. When the convex hulls do intersect, NPP has to be stated in terms of reduced convex hulls (RCHs), made up of convex pattern combinations whose coefficients are bound by a mo1 value and that are disjoint for suitable m. The GSK and MDM methods have recently been extended to solve NPP for RCHs using the particular structure of the extreme points of a RCH. While effective, their reliance on extreme points may make them computationally costly, particularly when applied in a kernel setting. In this work we propose an alternative clipped extension of classical MDM that results in a simpler algorithm with the same classification accuracy than that of the extensions already mentioned, but also with a much faster numerical convergence.

The nearest point problem (NPP), i.e., finding the closest points between two disjoint convex hulls, has two classical solutions, the Gilbert-Schlesinger-Kozinec (GSK) and Mitchell-Dem'yanov-Malozemov (MDM) algorithms. When the convex hulls do intersect, NPP has to be stated in terms of reduced convex hulls (RCHs), made up of convex pattern combinations whose coefficients are bound by a mo1 value and that are disjoint for suitable m. The GSK and MDM methods have recently been extended to solve NPP for RCHs using the particular structure of the extreme points of a RCH. While effective, their reliance on extreme points may make them computationally costly, particularly when applied in a kernel setting. In this work we propose an alternative clipped extension of classical MDM that results in a simpler algorithm with the same classification accuracy than that of the extensions already mentioned, but also with a much faster numerical convergence.

Microblogging sites allowing disseminating distasteful content. This has become vigorous and nearly unstoppable now. Spreading online fake news has been identified as one of the major top concern of online abuse. Due to the difficulty in preventing and evaluating what does fake news contain prior to publishing it online, if an algorithm is known for detecting fake news, then spreading online fake news wouldn’t exist in the first place, lead this paper to presents an evaluation of the effectiveness of algorithm(s), able to detect and filter to reasonable degree of accuracy what constitute an online fake news. The proposed approach is a multi-layered evaluations technique to be built as an app, where all information read online is associated with a tag, given a description of the facts about the contain. A proof of concept is provided for better understanding of the proposed techniques. This has contributed in providing possible steps to be taken by some popular Microblogging sites to...

Internet provides the facility to find customers worldwide without limitation use e-commerce market is effective. As a result the number of customers that rely on the Internet in the purchase has increased dramatically. In the field of computer security, phishing is a criminal activity that is trying to get sensitive information illegally. Sensitive information could include usernames, passwords and credit card details. This study aims to select the features or attributes in order to obtain the most influential attributes in detecting phishing websites. Selection feature using the correlation-based feature selection. Some of the most important attributes will be selected using the CFS and is applied to the binary logistic regression algorithms. Based on the research results show that CFS is able to eliminate redundant attributes. The subset of attributes generated have this level of accuracy is not much different from the full attributes. This level of accuracy before the selection ...

This paper studies the application of support vector machines (SVMs) to the detection and classification of rolling-element bearing faults. The training of the SVMs is carried out using the sequential minimal optimization (SMO) algorithm. In this paper, a mechanism for selecting adequate training parameters is proposed. This proposal makes the classification procedure fast and effective. Various scenarios are examined using two sets of vibration data, and the results are compared with those available in the literature that are relevant to this investigation.

This paper studies the application of support vector machines (SVMs) to the detection and classification of rolling-element bearing faults. The training of the SVMs is carried out using the sequential minimal optimization (SMO) algorithm. In this paper, a mechanism for selecting adequate training parameters is proposed. This proposal makes the classification procedure fast and effective. Various scenarios are examined using two sets of vibration data, and the results are compared with those available in the literature that are relevant to this investigation.

As most of human activities are being moved to cyberspace, phishers and other cybercriminals are making the cyberspace unsafe by causing serious risks to users and businesses as well as threatening global security and economy. Nowadays, phishers are constantly evolving new methods for luring user to reveal their sensitive information. To avoid falling victim to cybercriminals, a phishing detection algorithms is very necessary to be developed. Machine learning or data mining algorithms are used for phishing detection such as classification that categorized cyber users in to either malicious or safe users or regression that predicts the chance of being attacked by some cybercriminals in a given period of time. Many techniques have been proposed in the past for phishing detection but due to dynamic nature of some of the many phishing strategies employed by the cybercriminals, the quest for better solution is still on. In this paper, we propose a new phishing detection model based on Ex...

This paper presents a method for automatic classification of semantic relations between nominals using Sequential Minimal Optimization. We participated in the four categories of SEMEVAL task 4 (A: No Query, No Wordnet; B: Word-Net, No Query; C: Query, No WordNet; D: WordNet and Query) and for all training datasets. Best scores were achieved in category B using a set of feature vectors including lexical file numbers of nominals obtained from WordNet and a new feature WordNet Vector designed for the task 1 .

The identification of target proteins for diseased condition yields the development of the disease detection recommender system and drug discovery processes whose reticence can demolish the pathogen. The testing of this drug discovery is done through clinical and in addition through pre-clinical observations first on the creatures then on people. Thereafter the discovered drug is ready for public use. But if the drug discovery testing phase does not show the suitable consequences, then the entire task must be repeated. This repetitive clinical as well as the preclinical experimentation task is very cumbersome. But keeping in view the importance of the disease detection and drug discovery phase in protein identification as well as in the protein classification process this task must be done by researchers. The advancements in computational biology reveal the importance of computational prediction of protein function or to identify the target on the basis of protein sequence extracted...

Medical industry produces a significant portion of data whereas by adopting various Machine Learning models it can make accurate predictions about public healthcare that can be generalised. Transfer learning improves traditional machine learning by transferring the knowledge learned in one or more tasks and by using it for learning improvement in a related target task. In the current study, transfer learning with random forests was applied. Four datasets of medical interest obtained from the University of California, Irvine (UCI) Machine Learning Repository were used i.e., the BUPA-Liver Disease Dataset, the Breast Cancer Wisconsin Dataset, the Cleveland Heart Disease Dataset, and the Pima Indians Diabetes dataset. To our knowledge, there has been no study that applied Random Forests and Transfer Learning for these datasets. According to our results, our proposed method could provide significant accuracy rates in terms of diagnosing these disorders. Specifically, the classification ...

We propose a new generative model, and a new image similarity kernel based on a linked hierarchy of probabilistic segmentations. The model is used to efficiently segment multiple images into a consistent set of image regions. The segmentations are provided at several levels of granularity and links among them are automatically provided. Model training and inference in it is faster than most local feature extraction algorithms, and yet the provided image segmentation, and the segment matching among images provide a rich backdrop for image recognition, segmentation and registration tasks.

As most of human activities are being moved to cyberspace, phishers and other cybercriminals are making the cyberspace unsafe by causing serious risks to users and businesses as well as threatening global security and economy. Nowadays, phishers are constantly evolving new methods for luring user to reveal their sensitive information. To avoid falling victim to cybercriminals, a phishing detection algorithms is very necessary to be developed. Machine learning or data mining algorithms are used for phishing detection such as classification that categorized cyber users in to either malicious or safe users or regression that predicts the chance of being attacked by some cybercriminals in a given period of time. Many techniques have been proposed in the past for phishing detection but due to dynamic nature of some of the many phishing strategies employed by the cybercriminals, the quest for better solution is still on. In this paper, we propose a new phishing detection model based on Ex...

Microblogging sites allowing disseminating distasteful content. This has become vigorous and nearly unstoppable now. Spreading online fake news has been identified as one of the major top concern of online abuse. Due to the difficulty in preventing and evaluating what does fake news contain prior to publishing it online, if an algorithm is known for detecting fake news, then spreading online fake news wouldn’t exist in the first place, lead this paper to presents an evaluation of the effectiveness of algorithm(s), able to detect and filter to reasonable degree of accuracy what constitute an online fake news. The proposed approach is a multi-layered evaluations technique to be built as an app, where all information read online is associated with a tag, given a description of the facts about the contain. A proof of concept is provided for better understanding of the proposed techniques. This has contributed in providing possible steps to be taken by some popular Microblogging sites to...

As most of human activities are being moved to cyberspace, phishers and other cybercriminals are making the cyberspace unsafe by causing serious risks to users and businesses as well as threatening global security and economy. Nowadays, phishers are constantly evolving new methods for luring user to reveal their sensitive information. To avoid falling victim to cybercriminals, a phishing detection algorithms is very necessary to be developed. Machine learning or data mining algorithms are used for phishing detection such as classification that categorized cyber users in to either malicious or safe users or regression that predicts the chance of being attacked by some cybercriminals in a given period of time. Many techniques have been proposed in the past for phishing detection but due to dynamic nature of some of the many phishing strategies employed by the cybercriminals, the quest for better solution is still on. In this paper, we propose a new phishing detection model based on Ex...

Recent empirical evidence suggests that the Weston-Watkins support vector machine is among the best performing multiclass extensions of the binary SVM. Current state-of-the-art solvers repeatedly solve a particular subproblem approximately using an iterative strategy. In this work, we propose an algorithm that solves the subproblem exactly using a novel reparametrization of the Weston-Watkins dual problem. For linear WW-SVMs, our solver shows significant speed-up over the state-of-the-art solver when the number of classes is large. Our exact subproblem solver also allows us to prove linear convergence of the overall solver.

Recent empirical evidence suggests that the Weston-Watkins support vector machine is among the best performing multiclass extensions of the binary SVM. Current state-of-the-art solvers repeatedly solve a particular subproblem approximately using an iterative strategy. In this work, we propose an algorithm that solves the subproblem exactly using a novel reparametrization of the Weston-Watkins dual problem. For linear WW-SVMs, our solver shows significant speed-up over the state-of-the-art solver when the number of classes is large. Our exact subproblem solver also allows us to prove linear convergence of the overall solver.

In this study, an analysis was performed by examining the wind power potential of Kirklareli province which is in the west of Turkey. Statistical data between 2001 and 2007 was used in this study. The data was obtained from Kirklareli branch of State Meteorological Service. In Kirklareli region, wind speed forecasts regarding the year 2013 were made for windpower plants that are supposed to be built. WEKA tool was used for the performed analyzes. Algorithm which was used for forecasting is REPTree which is decision tree algorithm. There are two basic reasons to use REPTree algorithm. First, it produces better results compared to other machine learning methods, and secondly, the model produced with REPTree has a clear content. For this reason, new information can be gathered by using the tree model. This advantage of REPTree algorithm is combined with Bagging method and average model is generated by using the models produced by new training sets that are derived from the original tra...

As most of human activities are being moved to cyberspace, phishers and other cybercriminals are making the cyberspace unsafe by causing serious risks to users and businesses as well as threatening global security and economy. Nowadays, phishers are constantly evolving new methods for luring user to reveal their sensitive information. To avoid falling victim to cybercriminals, a phishing detection algorithms is very necessary to be developed. Machine learning or data mining algorithms are used for phishing detection such as classification that categorized cyber users in to either malicious or safe users or regression that predicts the chance of being attacked by some cybercriminals in a given period of time. Many techniques have been proposed in the past for phishing detection but due to dynamic nature of some of the many phishing strategies employed by the cybercriminals, the quest for better solution is still on. In this paper, we propose a new phishing detection model based on Ex...

In this study, an analysis was performed by examining the wind power potential of Kirklareli province which is in the west of Turkey. Statistical data between 2001 and 2007 was used in this study. The data was obtained from Kirklareli branch of State Meteorological Service. In Kirklareli region, wind speed forecasts regarding the year 2013 were made for windpower plants that are supposed to be built. WEKA tool was used for the performed analyzes. Algorithm which was used for forecasting is REPTree which is decision tree algorithm. There are two basic reasons to use REPTree algorithm. First, it produces better results compared to other machine learning methods, and secondly, the model produced with REPTree has a clear content. For this reason, new information can be gathered by using the tree model. This advantage of REPTree algorithm is combined with Bagging method and average model is generated by using the models produced by new training sets that are derived from the original tra...

Sistem deteksi kantuk dirancang menggunakan Elektrokardiogram (EKG) dengan Jaringan Saraf Tiruan Radial Basis Function dan Particle Swarm Optimization (JST RBF-PSO). Karolinska Sleepiness Scale (KSS) menjadi acuan tingkat kantuk yang dikelompokkan menjadi kelas terjaga dan kelas mengantuk. Sistem ini menggunakan algoritma Pan-Tomkins untuk menentukan interval RR dari EKG. Fitur yang digunakan adalah 15 parameter fitur statistik. Pelatihan dan pengujian data menggunakan JST RBF-PSO dengan metode validasi silang. PSO digunakan untuk mengoptimasi parameter utama JST RBF yaitu bobot, pusat dan lebar. Sistem deteksi kantuk ini diuji menggunakan DROZY Database. Hasil penelitian menunjukkan akurasi sistem ini pada segmentasi 40 detik, jumlah neuron 150 dan 15 fitur statistik sebesar 88,36%.

Nowadays, phone or smartphone and internet are something that cannot be separated with human. One of negative effects of using internet is cyber crime like a phishing url. Phishing url is usually used to collecting personal information like pin number, credit card etc. There are many type of classification algorithm, two of them is Naïve Bayes and C4.5. Both of the alogithm is good for recognizing a phishing url. This website created are used to classify a unkown url with Naïve Bayes and C4.5 algorithm. The accuracy of C4.5 algorithm is 87.11% and 78.48% for Naïve Bayes. The average time needed to processing one url is 21.78 second for C4.5 and 23.31 second for Naïve Bayes.

Kernel descriptors [1] provide a unified way to generate rich visual feature sets by turning pixel attributes into patch-level features, and yield impressive results on many object recognition tasks. However, best results with kernel descriptors are achieved using efficient match kernels in conjunction with nonlinear SVMs, which makes it impractical for large-scale problems. In this paper, we propose hierarchical kernel descriptors that apply kernel descriptors recursively to form image-level features and thus provide a conceptually simple and consistent way to generate imagelevel features from pixel attributes. More importantly, hierarchical kernel descriptors allow linear SVMs to yield stateof-the-art accuracy while being scalable to large datasets. They can also be naturally extended to extract features over depth images. We evaluate hierarchical kernel descriptors both on the CIFAR10 dataset and the new RGB-D Object Dataset consisting of segmented RGB and depth images of 300 everyday objects.

We hypothesize that machine-learning algorithms (MLA) can classify completer and simulated suicide notes as well as mental health professionals (MHP). Five MHPs classified 66 simulated or completer notes; MLAs were used for the same task. Results: MHPs were accurate 71% of the time; using the sequential minimization optimization algorithm (SMO) MLAs were accurate 78% of the time. There was no significant difference between the MLA and MPH classifiers. This is an important first step in developing an evidence based suicide predictor for emergency department use.

Web phishing is a type of cybercrime that occasionally threatens the online activities of website visitors. Web phishing uses a phoney website page that closely mimics the legitimate Website in order to fool its target into providing crucial information. Web phishing attacks also continue to grow in popularity year after year. As a result, it is vital to design a web phishing detection system in order to reduce the number of victims and financial losses caused by web phishing attacks. The development of a web phishing detection system continues to this day, with machine learning being the most often used model. Unfortunately, the construction of a machine learning-based web phishing detection system frequently employs only a single classification step; however, the feature selection process enables an increase in the performance of the resultant classification. Thus, an experiment was conducted in this paper by using a feature selection procedure based on the Pearson correlation alg...

Corporate credit rating analysis is one of the most important financial problems, which has attracted lots of research interests in the literature. Recent studies have shown that Artificial Intelligence (AI) methods can achieve better performance than traditional statistical methods. In this paper, Multilayer Perceptrons (MLPs) and multiple statistics methods are used to carry out multi-class credit rating of listed corporations in Tehran Stock Exchange (TSE). To accomplish this goal, a sample of these corporations is randomly selected and financial ratios as independent variables are computed for them. Dependent variable is a discrete variable of four groups as the corporation's bond ratings. The sample is split into training set and test set. With training set sample, the model is constructed through Multiple Discriminant Analysis, and with the remaining, hold out sample of corporations, the model is evaluated. Financial ratios for 126 manufacturing companies quoted in TSE have been used. Two models based on artificial neural networks and multiple discriminant analysis (MDA) are utilized to bond rating of corporations. ANN model achieved 81% and 68% accuracy rates in training and holdout samples, respectively. To evaluate the reliability of the model, the data were examined with multivariate discriminate analysis method. MDA reached 65.3% correct classifications of total initial sample. Results indicate that financial characteristics of corporations are good criterion for their bond rating and the same variables as other countries are significant in Iran too.

We hypothesize that machine-learning algorithms (MLA) can classify completer and simulated suicide notes as well as mental health professionals (MHP). Five MHPs classified 66 simulated or completer notes; MLAs were used for the same task. Results: MHPs were accurate 71% of the time; using the sequential minimization optimization algorithm (SMO) MLAs were accurate 78% of the time. There was no significant difference between the MLA and MPH classifiers. This is an important first step in developing an evidence based suicide predictor for emergency department use.

The linear instability of a solar pond containing porous material in the lower convective zone (LCZ) is investigated. It is found that, in general, for physically realistic values, solar ponds that contain porous material are more stable than ones that do not. Interestingly, it is found that the percentage of porous material contained in the LCZ does not stabilise the system in a linear fashion and, in particular, approximately 60% of porous material in the LCZ appears to optimise the maximum temperature that can be stored in the LCZ provided the proportion of heat flux extracted at the base of the pond, f, is held fixed at a physically realistic value of f ¼ 0:5. When f is varied it is found that there is a critical value of f = 0.6 below which inclusion of porous material stabilises the system but above which it does not.