Search

This paper introduces a novel thresholding and morphology-based skull stripping method for different MRI modalities. The proposed method is designed in a way which is easy to use and generates satisfactory results with minimal parameter adjustments. The method is evaluated on three different benchmark datasets and compared with nine state-of-the-art skull stripping methods. The experimental results suggest that the proposed method generates comparable results to some of the best methods in literature. However, unlike many other methods, it works well on different types of MRI scans. Moreover, this method generates the skull mask along with the brain mask that can be used to study various skull pathologies. [ABSTRACT FROM AUTHOR]

Due to advancements in data collection, storage, and processing techniques, machine learning has become a thriving and dominant paradigm. However, one of its main shortcomings is that the classical machine learning paradigm acts in isolation without utilizing the knowledge gained through learning from related tasks in the past. To circumvent this, the concept of Lifelong Machine Learning (LML) has been proposed, with the goal of mimicking how humans learn and acquire cognition. Human learning research has revealed that the brain connects previously learned information while learning new information from a single or small number of examples. Similarly, an LML system continually learns by storing and applying acquired information. Starting with an analysis of how the human brain learns, this paper shows that the LML framework shares a functional structure with the brain when it comes to solving new problems using previously learned information. It also provides a description of the LML framework, emphasizing its similarities to human brain learning. It also provides citation graph generation and scientometric analysis algorithms for the LML literatures, including information about the datasets and evaluation metrics that have been used in the empirical evaluation of LML systems. Finally, it presents outstanding issues and possible future research directions in the field of LML. [ABSTRACT FROM AUTHOR]

As the online social network technology is gaining all time high popularity and usage, the malicious behavior and attacks of spammers are getting smarter and difficult to track. The newer spamming approaches using the social engineering concepts are making traditional spam and spammer detection techniques obsolete. Especially, content-based filtering of spam messages and spammer profiles in online social networks is becoming difficult. Newer approaches for spammer detection using topological features are gaining attention. Further, the evaluation of ensemble classifiers for detection of spammers over social networking behavior-based features is still in its infancy. In this paper, we present an ensemble learning method for online social network security by evaluating the performance of some basic ensemble classifiers over novel community-based social networking features of legitimate users and spammers in online social networks. The proposed method aims to identify topological and community-based features from users’ interaction network and uses popular classifier ensembles – bagging and boosting to identify spammers in online social networks. Experimental evaluation of the proposed method is done over a real-world data set with artificial spammers that follow a behavior as reported in earlier literature. The experimental results reveal that the identified features are highly discriminative to identify spammers in online social networks.

Due to existence of a huge amount of textual data either on the World Wide Web or in textual databases like PubMed, the development of novel automatic keyphrase extraction methods has emerged as one of the key research problems in recent past. Consequently, a number of machine learning techniques, mostly supervised, have been proposed to extract keyphrases from text documents. But, one of the main bottlenecks that hinders the success of such systems is the requirement of annotated corpora for training purpose. In this paper, we propose the design of a deep text mining system to identify keyphrases in text documents that are either unstructured or semi-structured in nature. The novelty of our system lies in its applicability on a single document, instead of demanding a collection of annotated texts for training, to identify keyphrases embedded within it. The proposed system applies parsing techniques to identify candidate phrases. After mapping the original set of candidate phrases into a low-dimensional space using Singular Value Decomposition (SVD), the Markov Clustering (MCL) technique is applied to cluster related sentences together. Finally, considering each cluster as a document, Latent Dirichlet Allocation (LDA) is applied to identify feasible keyphrases that are presented to users in non-increasing order of their relevance score values. The efficacy of the proposed system is established through experimentation on datasets from two different domains. On comparative evaluation, we found that the proposed system outperforms KEA and KEA that apply the supervised machine learning approach for automatic keyphrase extraction from text documents.

Mammalian Target of Rapamycin (mTOR) is a Ser/Thr protein kinase, and its role is integral to the autophagy pathway in cancer. Targeting mTOR for therapeutic interventions in cancer through autophagy pathway is challenging due to the dual roles of autophagy in tumor progression. The architecture of mTOR reveals two complexes – mTORC1 and mTORC2, each having multiple protein subunits. mTOR kinase inhibitors target the structurally and functionally similar catalytic subunits of both mTORC1 and mTORC2. In this paper, we have explored two different categories of molecular features – descriptors and fingerprints for developing predictive models using machine learning techniques. Random Forest variable importance measures and autoencoders are used to identify molecular descriptors and fingerprints, respectively. We have built various predictive models using identified features and their combination for predicting mTOR kinase inhibitors. Finally, the best model based on the Mathew correlation co-efficient value over the validation dataset is selected for screening kinase SARfari bioactivity dataset. In this study, we have identified twenty best performing descriptors for predicting mTOR kinase inhibitors. To the best of our knowledge, it is the first study on integrating traditional machine learning and deep learning-based approaches for feature extraction to predict mTOR kinase inhibitors. [ABSTRACT FROM AUTHOR]

Online Social Networks (OSNs) are witnessing sophisticated cyber threats, that are generally conducted using fake or compromised profiles. Automated agents (aka socialbots), a category of sophisticated and modern threat entities, are the native of the social media platforms and responsible for various modern weaponized information-related attacks, such as astroturfing, misinformation diffusion, and spamming. Detecting socialbots is a challenging and vital task due to their deceiving character of imitating human behavior. To this end, this paper presents an attention-aware deep neural network model, DeepSBD, for detecting socialbots on OSNs. The DeepSBD models users’ behavior using profile, temporal, activity, and content information. It jointly models OSN users’ behavior using Bidirectional Long Short Term Memory (BiLSTM) and Convolutional Neural Network (CNN) architectures. It models profile, temporal, and activity information as sequences, which are fed to a two-layers stacked BiLSTM, whereas content information is fed to a deep CNN. We have evaluated DeepSBD over five real-world benchmark datasets and found that it performs significantly better in comparison to the state-of-the-arts and baseline methods. We have also analyzed the efficacy of DeepSBD at different ratios of socialbots and benign users and found that an imbalanced dataset moderately affects the classification accuracy. Finally, we have analyzed the discrimination power of different behavioral components, and it is found that both profile characteristics and content behavior are most impactful, whereas diurnal temporal behavior is the least effective for detecting socialbots on OSNs. [ABSTRACT FROM AUTHOR]

In online social networks (OSNs), socialbots are responsible for various malicious activities, and they are mainly programmed to imitate human-behavior to bypass the existing detection systems. The socialbots are generally successful in their malicious intent due to the existence of OSN users who follow them and thereby increase their reputation in the network. Analysis of the socialbot networks and their users is vital to comprehend the socialbot problem from target users' perspective. In this paper, we present a machine learning-based approach for characterizing and detecting socialbot targets, i.e., users who are susceptible to be trapped by the socialbots. We model OSN users based on their identity and behavior information, representing the static and dynamic components of their personality. The proposed approach classifies socialbot targets into three categories viz. active, reactive, and inactive users. We evaluate the proposed approach using three classifiers over a dataset collected from a live socialbot injection experiment conducted on Twitter. We also present a comparative evaluation of the proposed approach with a state-of-the-art method and show that it performs significantly better. On feature ablation analysis, we found that network structure and user intention and personality related dynamic features are most discriminative, whereas static features show the least impact on the classification. Additionally, following rate, multimedia ratio, and follower rate are most relevant to segregate different categories of the socialbot targets. We also perform a detailed topical and behavioral analysis of socialbot targets and found active users to be suspicious. Further, joy and agreeableness are the most dominating personality traits among the three categories of the users. [ABSTRACT FROM AUTHOR]

Online Social Networks (OSNs) are the modern communication media at the peak of Internet technology that facilitate users' ability to connect with friends and celebrities, and to disseminate breaking news and updates on real-life events. Connections and interactions among OSN users generate a massive amount of data containing a rich collection of knowledge that could be useful to various realworld data modeling and predictive analytics problems like open-source intelligence, business intelligence, event prediction, and product recommendations. On the Web, adversaries explore and target products and services for misuse and vulnerabilities, and OSNs are no exception. Their large user base, easy-to-use functionality, and open nature further attract antisocial elements to these OSNs. In OSNs, cybercrimes and illicit activities are generally committed using various forms of fake profiles, such as clone profiles,1 sybil,2 and bots. [ABSTRACT FROM AUTHOR]

Twitter is a popular microblogging platform, which facilitates users to express views and thoughts on day-to-day events using short texts limited to a maximum of 280 characters. However, it is generally targeted by socialbots for political astroturfing, advertising, spamming, and other illicit activities due to its open and real-time information sharing and dissemination nature. In this paper, we present a socialbots analysis-driven graph-based approach for identifying coordinated campaigns among Twitter users. To this end, we first present statistical insights derived from the analysis of logged data of 98 socialbots, which were injected in Twitter and associated with top-six Twitter using countries. In the analysis, we study and present the impact of socialbots' profile features, such as age and gender on infiltration. We also present a multi-attributed graph-based approach to model the profile attributes and interaction behavior of users as a similarity graph for identifying different groups of synchronized users involved in coordinated campaigns. The proposed approach is experimentally evaluated using four different evaluation parameters over a real dataset containing socialbots' trapped user profiles. The evaluation of identified campaigns in the form of clusters reveals the traces of spammers, botnets, and other malicious users. [ABSTRACT FROM AUTHOR]

Community detection from networks is one of the long standing and challenging tasks in the field of complex network research. Detection of communities poses numerous challenges in terms of their overlapping and hierarchical nature, dynamics of networks and underlying communities, scalability of detection algorithms on large scale networks to mention a few. Traditional community detection methods are not readily scalable to large networks mainly due to the computation of global network metrics. This paper presents a novel scalable overlapping community detection approach for large scale networks by presenting a MapReduce framework based implementation of a density-based local community detection method. The method is divided in two stages where the first stage uses a MapReduce approach to identify a mutual-core connected subgraph of the underlying network. The second stage uses an existing connected component detection method, implemented via MapReduce, to identify connected components in the mutual-core connected subgraph generated in the first stage. A community is then taken as the union of the core-nodes in a connected component and the respective density-based neighborhood of each core-node in the connected component. The resulting approach is among the first scalable overlapping community detection methods proposed in literature. [ABSTRACT FROM AUTHOR]

The frequent usage of figurative language on online social networks, especially on Twitter, has the potential to mislead traditional sentiment analysis and recommender systems. Due to the extensive use of slangs, bashes, flames, and non-literal texts, tweets are a great source of figurative language, such as sarcasm, irony, metaphor, simile, hyperbole, humor, and satire. Starting with a brief introduction of figurative language and its various categories, this article presents an in-depth survey of the state-of-the-art techniques for computational detection of seven different figurative language categories, mainly on Twitter. For each figurative language category, we present details about the characterizing features, datasets, and state-of-the-art computational detection approaches. Finally, we discuss open challenges and future directions of research for each figurative language category. [ABSTRACT FROM AUTHOR]

Due to proliferation of competitive online Business-to-Consumer (B2C) models, it is becoming a challenging task for new users to choose best products, based on existing users' reviews residing on different e-commerce websites. On analysis, it is found that the opinions of the existing customers play an important role for new customers in making appropriate purchase decisions. Though there are some online websites that provide aggregation of basic product information from multiple sources, there is a negligible research effort in the direction of opinion-based product ranking. In this paper, we propose an Opinion-based Multi-Criteria Ranking (OMCR) approach, which amalgamates structural and content-based features of review documents to rank different alternatives of the online products. It uses a total number of five features based on reviews' meta-data and contents to rank different alternatives using multi-criteria decision making approaches. OMCR also incorporates a sentiment analysis and visualization approach to determine sentiment polarity values and visualize them in a comprehendible manner. Experiments are conducted over two different real datasets, and efficacy of OMCR is assessed using set intersection method, which is generally used to compare two ranked lists in terms of their overlapping score. [ABSTRACT FROM AUTHOR]

Twitter is one of the most popular microblogging services, which is generally used to share news and updates through short messages restricted to 280 characters. However, its open nature and large user base are frequently exploited by automated spammers, content polluters, and other ill-intended users to commit various cybercrimes, such as cyberbullying, trolling, rumor dissemination, and stalking. Accordingly, a number of approaches have been proposed by researchers to address these problems. However, most of these approaches are based on user characterization and completely disregarding mutual interactions. In this paper, we present a hybrid approach for detecting automated spammers by amalgamating community-based features with other feature categories, namely metadata-, content-, and interaction-based features. The novelty of the proposed approach lies in the characterization of users based on their interactions with their followers given that a user can evade features that are related to his/her own activities, but evading those based on the followers is difficult. Nineteen different features, including six newly defined features and two redefined features, are identified for learning three classifiers, namely, random forest, decision tree, and Bayesian network, on a real dataset that comprises benign users and spammers. The discrimination power of different feature categories is also analyzed, and interaction- and community-based features are determined to be the most effective for spam detection, whereas metadata-based features are proven to be the least effective. [ABSTRACT FROM PUBLISHER]

The growing popularity of online social media is leading to its widespread use among the online community for various purposes. In the recent past, it has been found that the web is also being used as a tool by radical or extremist groups and users to practice several kinds of mischievous acts with concealed agendas and promote ideologies in a sophisticated manner. Some of the web forums are predominantly being used for open discussions on critical issues influenced by radical thoughts. The influential users dominate and influence the newly joined innocent users through their radical thoughts. This paper presents an application of collocation theory to identify radically influential users in web forums. The radicalness of a user is captured by a measure based on the degree of match of the commented posts with a threat list. Eleven different collocation metrics are formulated to identify the association among users, and they are finally embedded in a customized PageRank algorithm to generate a ranked list of radically influential users. The experiments are conducted on a standard data set provided for a challenge at ISI-KDD’12 workshop to find radical and infectious threads, members, postings, ideas, and ideologies. Experimental results show that our proposed method outperforms the existing UserRank algorithm. We also found that the collocation theory is more effective to deal with such ranking problem than the textual and temporal similarity-based measures studied earlier. [ABSTRACT FROM PUBLISHER]

Social network analysis ( SNA) is a multidisciplinary field dedicated to the analysis and modeling of relations and diffusion processes among various objects in nature and society, and other information/knowledge processing entities with an aim of understanding how the behavior of individuals and their interactions translates into large-scale social phenomenon. Because of exploding popularity of online social networks and availability of huge amount of user-generated content, there is a great opportunity to analyze social networks and their dynamics at resolutions and levels not seen before. This has resulted in a significant increase in research literature at the intersection of the computing and social sciences leading to several techniques for social network modeling and analysis in the area of machine learning and data mining. Some of the current challenges in the analysis of large-scale social network data include social network modeling and representation, link mining, sentiment analysis, semantic SNA, information diffusion, viral marketing, and influential node mining. WIREs Data Mining Knowl Discov 2013, 3:408-444. doi: 10.1002/widm.1105 Conflict of interest: The authors have declared no conflicts of interest for this article. For further resources related to this article, please visit the WIREs website. [ABSTRACT FROM AUTHOR]

Word based searches for relevant information from texts retrieve a huge collection and burden the user with information overload. Ontology based text information retrieval can perform concept-based search and extract only relevant portions of text containing concepts that are present in the query or those that are semantically linked to query concepts. While these systems have better precision of retrieval than general-purpose search engines, problems arise with those domains where ontological concepts cannot be unambiguously described using precise property descriptors. Besides, the ontological descriptors may not exactly match text descriptions or the user given descriptors in query. In such situations, uncertainty based reasoning principles can be applied to find approximate matches to user queries. In this paper we have presented a framework to enhance traditional ontological structures with fuzzy descriptors. The fuzzy ontology structure has been used to locate and extract both precise and imprecise descriptions of concepts from Web documents and then store them in a structured knowledge base. The design of the structured knowledge base, which in our case is a database, is also derived from the underlying fuzzy ontology representing the domain. User queries are processed in two stages. In the first stage, precise SQL queries are formulated and processed over the knowledge base to find a possible answer set. In the second stage, fuzzy reasoning is applied to compute the relevance of the answers in the answer set with respect to the query. We have provided experimental validation of the approach through knowledge-extraction and query processing executed over a diverse set of domains. [ABSTRACT FROM AUTHOR]

In this paper, we propose a unified framework, HOCTracker, for tracking the evolution of hierarchical and overlapping communities in online social networks. Unlike most of the dynamic community detection methods, HOCTracker adapts a preliminary community structure towards dynamic changes in social networks using a novel density-based approach for detecting overlapping community structures, and automatically tracks evolutionary events like birth, growth, contraction, merge, split, and death of communities. It uses a novel and efficient log-based approach to map evolutionary relations between communities identified at two consecutive time-steps of a dynamic network, which considerably reduces the number of community comparisons. Moreover, it does not require an ageing function to remove old interactions for identifying community evolutionary events. HOCTracker is applicable to diirected/undirected and weighted/unweighted networks. Experimental results have shown that community structures identified by HOCTracker on some well-known benchmark networks are significant and in general better that the community structures identified by the state-of-the-art methods. [ABSTRACT FROM PUBLISHER]

Owing to increasing size of the real-world networks, their processing using classical techniques has become infeasible. The amount of storage and central processing unit time required for processing large networks is far beyond the capabilities of a high-end computing machine. Moreover, real-world network data are generally distributed in nature because they are collected and stored on distributed platforms. This has popularized the use of the MapReduce, a distributed data processing framework, for analyzing real-world network data. Existing MapReduce-based methods for connected components detection mainly struggle to minimize the number of MapReduce rounds and the amount of data generated and forwarded to the subsequent rounds. This article presents an efficient MapReduce-based approach for finding connected components, which does not forward the complete set of connected components to the subsequent rounds; instead, it writes them to the Hadoop Distributed File System as soon as they are found to reduce the amount of data forwarded to the subsequent rounds. It also presents an application of the proposed method in contact tracing. The proposed method is evaluated on several network data sets and compared with two state-of-the-art methods. The empirical results reveal that the proposed method performs significantly better and is scalable to find connected components in large-scale networks.