Concept Detection in Multimedia Web Resources About Home Made Explosives

We propose a technique for video shot segmentation and classification based on singular value decomposition (SVD). For the input video sequence, we create a feature-frame matrix A, and perform the SVD on it. From this SVD, we are able to not only derive the refined feature space to better segment the video sequence along the time axis, but also define metrics to enable classifications of the detected video shots. Using these SVD properties, we achieve the two goals of accurate video shot segmentation, and visual content-based shot classification at the same time.

Far-right–inspired violence in the United States is on the rise (Perliger, 2013), and there is a growing realization within policy circles of the increasing threat from so-called "lone wolf" terrorists of different ideological hues (Simon, 2013). A more in-depth understanding of individual offenders and their behaviors (both antecedent and offense-commission) is therefore of great importance not only to the academic literature but also for practitioners tasked with countering the problem. The need for such research is also driven by the fact that a preponderance of studies have focused on violent Islamist organizations for more than a decade. Given the nature of the threat presented by al-Qaeda, this is understandable, but a side effect is that our understanding of the nature of far-right–inspired homicide lags behind our understanding of the behaviors associated with analogous movements. Recognizing these features of the existing literature and the current security environment, Gruenewald, Chermak, and Freilich (2013, this issue) provide an empirical assessment of the behavioral underpinnings of far-right–inspired homicide offenders and compare group-based offenders with lone offenders. Gruenewald et al. (2013) test 23 hypotheses, each of which is derived from a careful examination of the existing literature. Using inferential statistics, they find several significant differences between lone and group offenders. For example, the bivariate analysis suggests that compared with group offenders, far-right lone offenders are statistically more likely to be unmarried, have experience of the military, live alone, use firearms, select government targets, kill more people, and die in the commission of their offense. The multivariate analysis also suggests that age and mental illness also significantly distinguish lone offenders from group offenders (e.g., lone offenders are typically older and more likely to have a history of mental illness). For the remainder of this essay, we consider the salient issues in the available relevant literature in relation to Gruenewald et al.'s (2013) study by identifying and discussing specific and actionable next policy and research steps, as well as this study's potential impact on the current state of knowledge. Reviews of the terrorism research literature have regularly highlighted the paucity of original data that inform analyses (Schmid and Jongman, 1988; Silke, 2001, 2004). In the most recent review of the literature, Silke (2013: 34) noted: "[O]ne feels that a great deal more needs to be done before research is consistently building on past work rather than rehashing old data." Although the use of inferential statistical methods is certainly on the rise with relation to terrorism research, these studies typically tend to be based on terrorist events rather than on the individual offenders. This is largely because of the data available from the Global Terrorism Database. More recently however, there have been a handful of large-N studies of terrorist group membership that primarily use descriptive statistics (Gill and Horgan, in press; Horgan and Morrison, 2011; Magouirk, Atran, and Sageman, 2008; Reinares, 2004; Sageman, 2004). Gruenewald et al.'s (2013) article is a major next step in the study of the individual terrorist. Their development of the U.S. Extremist Crime Database (ECDB) provides a solid empirical basis from which inferential statistical approaches to the study of extremists and most importantly extremist behavior can be conducted. Gruenewald et al.'s (2013) study highlights significant behavioral and personal differences between far-right–inspired lone and group offenders. Although some of these differences were expected based on their survey of research on lone-actor terrorism, they neglected to mention that a few studies of different crimes have found broadly similar results. For example, Alarid, Buron, and Hochstetler's (2009) study of robberies showed that group offenders were more likely to be younger, unmarried, and less educated. In terms of target selection, there was little difference between the difficulty levels of robberies chosen by group and solo offenders, which runs counter to much theorizing concerning "strength in numbers," and offending (see Felson, Baumer, and Messner, 2000, for example). Using vastly different samples, studies comparing lone and group rapists have found that the former tended to be significantly older (Bijleveld and Hendriks, 2003; Hauffe and Porter, 2009; Ullman, 1999; Wright and West, 1981), significantly more likely to be married and have children, significantly less likely to have a history of substance abuse (Hauffe and Porter, 2009), significantly more likely to have previous convictions and problematic personality structures (Bijleveld and Hendriks, 2003), and significantly more likely to be employed (Ullman, 1999). Hickle and Roe-Sepowitz's (2010) study of female juvenile arsonists found solo offenders more often came from unstable homes; more often experienced school difficulties, behavioral problems, and negative emotions; and expressed suicidal thoughts more regularly (the latter finding is particularly relevant to Gruenewald et al.'s study when we consider their finding that lone offenders were more likely to die in the commission of their offense). Block's (1985, 1986) study of Chicago-based homicides illustrated that group offenders were significantly younger than lone offenders. Cheatwood and Block's (1990) study on homicides in Baltimore also found a disproportionate number of juvenile offenders engaging in homicides as members of groups rather than as individuals. In the most comprehensive of such analyses, Clark (1995) analyzed data across nine U.S. cities. Again, group offenders were significantly younger. Similar to Hauffe and Porter's (2009) results (discussed previously), lone offenders were less likely to have a history of substance abuse than group offenders. Unlike research on robberies, group homicide offenders were more likely to have previous criminal convictions (Clark, 1995). Aggregating a large number of crimes, Van Mastrigt and Farrington (2011) analyzed U.K. police data; group offenders were significantly younger than lone offenders and tended to have more criminal convictions. This finding is common when the offender sample is aggregated across crime types (Erickson, 1971; Hindelang, 1976; Reiss and Farrington, 1991). Collectively, these studies should raise future questions about what factors drive these differences across crime types. For example, answers may lie in individual decision making, gatekeeper selection effects, and group dynamics and influences. The collective findings also suggest that some factors may be crime specific (e.g., previous convictions), whereas others may be crime general (e.g., age and mental illness). These studies also highlight, from a policy and practitioner perspective, whether desistance and disengagement from crime is a different process for solo and group offenders. The importance of Gruenewald et al.'s (2013) study lies not only in its use of inferential statistical methods and in its delineation between far-right–inspired lone and group offenders, but also the findings have important implications for issues long debated within the literature, particularly issues concerning mental illness and terrorist involvement, as well as issues concerned with group processes. In the 1970s, the first wave of studies on the individual terrorist offered psychopathological interpretations that perceived terrorist behavior to be essentially deviant. An underlying assumption in this research strand is that compulsion to join the militant organization, or the vulnerability to recruitment, is inherent in those who become engaged in militancy. Motivation to participate is therefore situated within the individual in isolation. To this end, researchers have postulated deviant characteristics of the terrorist. For Victoroff (2005), this field was "largely characterized by theoretical speculation based on subjective interpretation of anecdotal observations" (p. 3). In the 1980s, psychoanalytic approaches (with a particular emphasis on the concept of narcissism) largely replaced psychopathological explanations. These findings are reasonably similar to the preceding assumptions of the terrorist personality being essentially abnormal but perceive this abnormality as determined by unconscious drives and urges originating from childhood rather than inherent personality flaws. Over the past 15 years or so, a growing awareness of the need to understand the group process has come to fruition. This has largely been aided by empirical studies that have engaged in first-hand interviews and by the incorporation of theory from social psychology. This stream of research has been successful at downplaying the role of personality and mental health issues with regard to terrorist involvement to the extent that one noted academic, who in the past championed psychoanalytical interpretations of terrorist motivation, argued that "a clear consensus exists that it is not individual psychology, but group, organizational and social psychology, that provides the greatest analytical power in understanding this complex phenomenon" (Post, 2005: 7). Although the studies of the 1970s and 1980s were correctly debunked on methodological grounds, Gruenewald et al.'s (2013) findings suggest that the existing consensus that mental illness never plays a role in terrorism and that terrorist behavior can be purely explained by group processes is likely to be a fallacy. With regard to group processes, Gruenewald et al.'s (2013) study supports recent work that has started to reexamine the role of group processes in terrorism. For example, the debate has been widened by Taylor (2010: 121), who argued that "group-focused causal accounts fail to offer sufficient explanations of the behaviors identified" in his case studies, whereas Gill (2012) attempted to refine the nature of role-specific qualities and group-based behavior. Although many studies of the individual terrorist are fixated with understanding motive and/or radicalization (this is also true for media accounts if we examine the coverage regarding the recent Boston Marathon bombings), Gruenewald et al. instead focus on behavior. Their findings that lone offenders are significantly more likely to have experience of the military and live alone also importantly provide insight into capability. Group membership provides individuals not only the operational capabilities necessary to perform a terrorist attack but also the mechanism for moral disengagement (Gill, 2012). Research literature from military psychology has suggested that many factors inhibit individuals from engaging in fatal violence against others. For example, Marshall concluded that "the average and healthy individual … has such an inner and usually unrealized resistance towards killing a fellow man that he will not of his own volition take life if it is possible to turn away from that responsibility … at the vital point" (cited in Grossman, 1995: 29). Absent a group setting, terrorist violence may be difficult to commit for practical and psychological reasons. In the absence of a group, an individual may need other factors to (a) provide security, (b) help overinhibitions against committing violence, (c) help develop the necessary practical skills, and (d) finance an operation. Here, variables found to be significant in the current study such as living alone at the time of the attack and military training provide the capabilities to turn motivation into action for lone offenders. This is also true for hypotheses that were unsupported, such as lone offenders being more likely to be unemployed. Without a group to help an individual plan, prepare, and carry out a successful attack, the lone offender must have a steady source of income. This observation is highlighted in the case of Anders Breivik in Norway, who financed his operations through several legal and illegal activities in the years before his attack and avoided delaying his attack further because he was having liquidity issues and feared he could not (a) keep renting the farm where he built his improvised explosive device and (b) rent a car for the operation itself. With regard to mental illness and terrorism, Gruenewald et al.'s (2013) finding that more than 40% of lone offenders had previously been diagnosed with mental illness is one of the most statistically significant findings from the multivariate analysis. Despite its importance, this finding receives minimal attention in the concluding remarks and policy implications. This may be born out of discrepancies across the literature concerning mental illness and terrorism, leading to the current existing false dichotomy that an act of targeted violence was either the actions of a terrorist or the actions of someone who was mentally ill. Gruenewald et al. do suggest important methods for identifying diagnoses, but we hold the opinion that future research should use mental illness as a central unit of analysis. In doing so, it would provide a better understanding of the complexity of mental illness with relation to terrorist behavior and provide the possibility to examine how the other variables would affect and be affected by the offender suffering from mental illness. In other words, the mental illness variable is likely to stem from several other variables accounted for in the ECDB database and impact many other variables also. For example, research by Elisha, Castle, and Hocking (2006) illustrated that social isolation is highly prevalent among those diagnosed with a mental illness. Of their 3,800 observations of individuals diagnosed with a psychotic illness, 58% were also characterized as being socially withdrawn. Butterworth and Rodgers (2008) demonstrated that couples with at least one spouse diagnosed with a mental illness suffer higher rates of marital disruption than non–mentally-ill couples. Using mental illness as a central variable with relation to the criminogenic factors may also be a promising approach. As Anderson (1997) explained, the symbiotic relationship between mental illness and criminal behavior is complex. Comorbidity of mental illness with other behaviors is well documented across the literature and includes substance abuse (Todd et al., 2004) and violent and criminal convictions (Anderson, 1997). It is also a common understanding in psychiatric practice (Diagnostic and Statistical Manual [DSM] and International Classification of Diseases [ICD]). Gruenewald et al.'s (2013) study found that 34% of lone offenders had substance abuse problems and that 61.7% had prior arrests. Taking into consideration the complexities that comorbidity brings, it may be necessary in the future to distinguish between lone offenders who were mentally ill and those who were not to determine and compare the rates of substance abuse and criminal convictions, as well as compare these with group actors. Many different mental illnesses display impulsivity as a problem behavior (DSM and ICD), which affects many aspects of living and may affect the attack types of offenders, as well as the primary weapon usage. The current study found that 84.8% of lone offenders used a gun, compared with 40.2% of group offenders. This finding is highly significant, but if one takes into account the country that the data were taken from (the United States) and the gun control laws, it could be hypothesized that lone offenders were more likely to use a gun because nearly half struggled with impulse control and guns are freely and immediately available. It is also interesting to note the potential link between the highly significant finding that lone offenders were more likely to partake in a suicide mission compared with group offenders given the higher levels of mental illness found in lone actors. This finding is interesting because suicide risks are much higher in the mentally ill population, and if non–mentally-ill lone actors were investigated separately, this figure may decrease to a value approaching that of group-based offenders. Finally in relation to mental illness, Gill, Horgan, and Deckert (in press), and many of the group versus solo offender studies cited previously, also found the average age of lone offenders to be older than that of group offenders. This finding may provide support for the idea that a different temporal trajectory into criminality is at play for lone- and group-based offenders. It may be necessary to consider and examine whether the increased age range is a result of the later onset of mental illness. For example, military service personnel are at high risk of developing posttraumatic stress disorder (PTSD) (Sundin et al., 2010), which again may in part explain the highly significant discrepancies between lone and group offenders Gruenewald et al. (2013) identified three policy recommendations that should be tailored from preexisting policies to fit the problem of lone offenders. Importantly, Gruenewald et al. make the point that lone offender attacks are preventable, which runs counter to several current studies that have not had the benefit of the high-quality data sets or rigorous methodologies that Gruenewald et al. used (Barnes, 2012; Michael, 2012; Simon, 2013). The first recommendation is to increase intelligence partnerships between law enforcement and other organizations such as military intelligence units and social service agencies. This suggestion stems from the finding that lone offenders were significantly more likely to have military experience and a history of mental illness. We suggest that this could go further. The policy implications of our suggestions regarding the future study of mental illness with relation to terrorism suggests using multiple agencies at numerous intervention stages (social services, policing agencies, the courts, military organizations, psychiatric hospitals, weapons providers, community programs, financial agencies, intelligence agencies, and the media), which could all cooperate to aid in prevention and reduction efforts. The second recommendation involves increasing technological surveillance, which spans a wide range of functions including monitoring websites and chat rooms, collecting intelligence, creating fake websites, building relationships with chat room moderators (and thereby increasing guardianship), and countering extremist narratives. These behaviors, if applied correctly, could also help prevent and disrupt online group-based offenders. The final recommendation is to sharpen threat assessment tools, which is similar to recommendations stemming from broadly similar studies of school shooters and politically motivated assassins (Fein and Vossekuil, 1999; Vossekuil et al., 2002). The significant behavioral differences between lone and group offenders found using the ECDB data suggest the need for a tailored response to both subgroups. It also serves to raise further questions about whether these differences hold true for other ideological motivations. Gill, Horgan, and Deckert's (in press) study suggests that identifiable behavioral differences exist among al-Qaeda–inspired, extreme right-wing, and single-issue lone-actor terrorists. The ECDB data also focus solely on offenders who commit fatal attacks. It would be interesting to observe what patterns hold for those who committed attacks that caused no fatalities as well as those whose plots were interrupted preattack. Finally, as mentioned, the ECDB data focus on only U.S.-based offenders, which leads to the question of whether this group of offenders is more likely to use firearms because of the smaller opportunity costs involved. In conclusion, hopefully the highly important results found in Gruenewald et al.'s (2013) study will lead to a greater comparative examination of these issues across ideological domains and geographic boundaries. Paul Gill is a lecturer at University College London's Department of Security and Crime Science. His research focuses upon the behavioral underpinnings of terrorist attacks. Emily Corner is a research student at University College London's Department of Security and Crime Science. Her dissertation examines the nature of mental illness in a population of 119 lone-actor terrorists.

Why is terrorism such a slippery concept? Unlike other concepts employed in studies of conflict, terrorism is particularly hard to pin down. Can terrorism be equated with civilian victimization? Is it a refined form of coercive violence, based on the spread of terror? Despite the endless terminological debate, there is an underlying logic in the way in which we speak about terrorism. This book is aimed at uncovering this logic, providing a systematic, coherent, and rigorous conceptualization of the phenomenon. Terrorism is a unique form of political violence due to the underground nature of its violence. Unlike other forms of violence, underground political violence does not require any control of the territory in which the attack takes place. The quintessential underground attack is an improvised explosive device in the enemy’s terrain. Because terrorist attacks do not presuppose territorial control, they tend to be ephemeral. Understanding terrorism as underground political violence makes sense of many traits that are commonly associated with the concept: its international dimension, the possibility of lone-actors, and state terrorism as covert operations abroad. Terrorism tends to emerge in highly asymmetric conditions and is frequently used by the weaker party (hence the motto that terrorism is the weapon of the weak). Through large-n empirical tests and case studies (ETA, Tupamaros, the Shining Path, Fatah, ISIS, and the Palestinian insurgent groups), this book shows that the conception of terrorism as underground political violence has some ‘organizing’ power and can be used in empirical analysis to shed light on the determinants and dynamics of terrorism.

To achieve more efficient video indexing and access, we introduce a video database management framework and strategies for video content structure and events mining. The video shot segmentation and representative frame selection strategy are first utilized to parse the continuous video stream into physical units. Video shot grouping, group merging, and scene clustering schemes are then proposed to organize the video shots into a hierarchical structure using clustered scenes, scenes, groups, and shots, in increasing granularity from top to bottom. Then, audio and video processing techniques are integrated to mine event information, such as dialog, presentation and clinical operation, from the detected scenes. Finally, the acquired video content structure and events are integrated to construct a scalable video skimming tool which can be used to visualize the video content hierarchy and event information for efficient access. Experimental results are also presented to evaluate the performance of the proposed framework and algorithms.

Optimized deep learning enabled lecture audio video summarization

Growth in system complexity increases the need for automated techniques dedicated to different log analysis tasks such as Log-based Anomaly Detection (LAD). The latter has been widely addressed in the literature, mostly by means of a variety of deep learning techniques. However, despite their many advantages, that focus on deep learning techniques is somewhat arbitrary as traditional Machine Learning (ML) techniques may perform well in many cases, depending on the context and datasets. In the same vein, semi-supervised techniques deserve the same attention as supervised techniques since the former have clear practical advantages. Further, current evaluations mostly rely on the assessment of detection accuracy. However, this is not enough to decide whether or not a specific ML technique is suitable to address the LAD problem in a given context. Other aspects to consider include training and prediction times as well as the sensitivity to hyperparameter tuning, which in practice matters to engineers. In this paper, we present a comprehensive empirical study, in which we evaluate a wide array of supervised and semi-supervised, traditional and deep ML techniques w.r.t. four evaluation criteria: detection accuracy, time performance, sensitivity of detection accuracy and time performance to hyperparameter tuning. Our goal is to provide much stronger and comprehensive evidence regarding the relative advantages and drawbacks of alternative techniques for LAD. The experimental results show that supervised traditional and deep ML techniques fare similarly in terms of their detection accuracy and prediction time on most of the benchmark datasets considered in our study. Moreover, overall, sensitivity analysis to hyperparameter tuning with respect to detection accuracy shows that supervised traditional ML techniques are less sensitive than deep learning techniques. Further, semi-supervised techniques yield significantly worse detection accuracy than supervised techniques.

The explosive growth of online short videos has brought great challenges to the efficient management of video content classification, retrieval, and recommendation. Video features for video management can be extracted from video image frames by various algorithms, and they have been proven to be effective in the video classification of sensor systems. However, frame-by-frame processing of video image frames not only requires huge computing power, but also classification algorithms based on a single modality of video features cannot meet the accuracy requirements in specific scenarios. In response to these concerns, we introduce a short video categorization architecture centered around cross-modal fusion in visual sensor systems which jointly utilizes video features and text features to classify short videos, avoiding processing a large number of image frames during classification. Firstly, the image space is extended to three-dimensional space–time by a self-attention mechanism, and a series of patches are extracted from a single image frame. Each patch is linearly mapped into the embedding layer of the Timesformer network and augmented with positional information to extract video features. Second, the text features of subtitles are extracted through the bidirectional encoder representation from the Transformers (BERT) pre-training model. Finally, cross-modal fusion is performed based on the extracted video and text features, resulting in improved accuracy for short video classification tasks. The outcomes of our experiments showcase a substantial superiority of our introduced classification framework compared to alternative baseline video classification methodologies. This framework can be applied in sensor systems for potential video classification.

Machine learning (ML) is being applied to a diverse and ever-growing set of domains. In many cases, domain experts - who often have no expertise in ML or data science - are asked to use ML predictions to make high-stakes decisions. Multiple ML usability challenges can appear as result, such as lack of user trust in the model, inability to reconcile human-ML disagreement, and ethical concerns about oversimplification of complex problems to a single algorithm output. In this paper, we investigate the ML usability challenges that present in the domain of child welfare screening through a series of collaborations with child welfare screeners. Following the iterative design process between the ML scientists, visualization researchers, and domain experts (child screeners), we first identified four key ML challenges and honed in on one promising explainable ML technique to address them (local factor contributions). Then we implemented and evaluated our visual analytics tool, Sibyl, to increase the interpretability and interactivity of local factor contributions. The effectiveness of our tool is demonstrated by two formal user studies with 12 non-expert participants and 13 expert participants respectively. Valuable feedback was collected, from which we composed a list of design implications as a useful guideline for researchers who aim to develop an interpretable and interactive visualization tool for ML prediction models deployed for child welfare screeners and other similar domain experts.

Short videos on the Internet are growing exponentially, and the number of videos uploaded every day is huge; people also involve a lot of video data in real life. People can retrieve and view all kinds of videos, but it also brings a lot of problems. On the one hand, the accumulation of a large number of videos makes people unable to find the videos they want quickly, and the repeated scenes in the videos will also waste people's time and energy; on the other hand, a large amount of video data also brings enormous pressure to storage. Aiming at the problems of inaccurate selection of key frames and how to select video frame features in existing video summarization models, this paper proposes a multi-feature-based video summarization generation model (DME-VSNet), which extracts multiple features of video frames. Including importance score, image memory strength and image entropy. Aiming at the problem of inaccurate video shot segmentation, this model proposes a video shot segmentation algorithm based on TransNet network, which divides the original video into several short shots through shot boundaries; the model inputs the above three features into the proposed The video frame score is obtained in the MLP architecture, and the key frame is selected by the score to generate a video summary. The effectiveness of the video shot segmentation method based on TransNet network and the overall model based on convolutional neural network is verified by comparative experiments. The experimental results show that the evaluation results of the video summaries generated by the three features are better.

Today, ICT, web resources and multimedia contents have become prevalent in Malaysian university classrooms; hence, the learning approaches need to be redesigned for enabling students to use these technologies in co-constructing new meaning. This study analyses student’s perception and their peer interaction in the constructivist–collaborative learning environment (CCLE) with multimedia contents and Web 2.0 tools, to address the research question: What is the impact of CCLE on student learning process? By using factor analysis on the questionnaire results, and applying content analysis on students’ comments in Web 2.0 tools and feedback given by 182 students at INTI International University, several components of constructivist–collaborative learning were identified. The discussion in this study focuses on “team” and its sub-components include group dynamics, students’ prior knowledge with team and group processing. The discussion provides more understanding on the impact on student learning process for enhancing the learning experiences in Malaysian university class environment.

Chronic Kidney Disease (CKD) are a universal issue for the well-being of people as they result in morbidities and deaths with the onset of additional diseases. Because there are no clear early symptoms of CKD, people frequently miss them. Timely identification of CKD allows individuals to acquire proper medications to prevent the development of the diseases. Machine learning technique (MLT) can strongly assist doctors in achieving this aim due to their rapid and precise determination capabilities. Many MLT encounter inappropriate features in most databases that might lower the classifier’s performance. Missing values are filled using K-Nearest Neighbor (KNN). Adaptive Weight Dynamic Butterfly Optimization Algorithm (AWDBOA) are nature-inspired feature selection (FS) techniques with good explorations, exploitations, convergences, and do not get trapped in local optimums. Operators used in Local Search Algorithm-Based Mutation (LSAM) and Butterfly Optimization Algorithm (BOA) which use diversity and generations of adaptive weights to features for enhancing FS are modified in this work. Simultaneously, an adaptive weight value is added for FS from the database. Following the identification of features, six MLT are used in classification tasks namely Logistic Regressions (LOG), Random Forest (RF), Support Vector Machine (SVM), KNNs, Naive Baye (NB), and Feed Forward Neural Network (FFNN). The CKD databases were retrieved from MLT repository of UCI (University of California, Irvine). Precision, Recall, F1-Score, Sensitivity, Specificity, and accuracy are compared to assess this work’s classification framework with existing approaches.

With the mixing of machine learning (ML) algorithms, stock market forecasting has become an area of interest for researchers. This paper presents a review of the literature on ML techniques for price prediction in Indian stock markets. It analyses major algorithms such as linear regression, support vector machines (SVM), random forest, and neural network, as well as discussing their theoretical, advantageous and disadvantageous aspects. This research offers an exhaustive assessment of the implementation of diverse ML techniques for predicting trends in the Indian stock market. The swift changes and the nonlinear nature of stock markets make the problem of prediction extremely difficult. The author demonstrates how ML techniques are used to improve the accuracy of predictions by analysing historical data, sentiments of the market, and various technical indicators. Furthermore, it discusses issues such as crash in data, unpredictability of the market, and overfitting that create challenges for forecasting in ML. This review is aimed at providing substantial evidence in the implementation of various ML techniques adjusted to the Indian stock market for predicting stock price movements and providing information where further research is needed.

A search for experts is a search for individuals with skills and knowledge in a particular field. Early studies studied profiling of experts in the expert search application by using keyword query. The use of concept-based query that we highlight is able to match experts and expertise to the searched field. Existing expert search systems perform search merely based on the experts' publication information. There are no relations to other information that can act as support resources to relate them to related fields. The technique to search for expert based on publication information narrows the search scope. There are other aspects that are related to the experts such as education, teachings, supervisions, consultations, conferences, community services and professional affiliations. The use of ontology that is based on concept can also help to expand the search results as well as to link to other relations. The purpose of this research is to design a framework for ontology development of Information, Communication and Technology (ICT) experts, to be used in the development of ICT expert directory. This framework is developed based on the critical evaluation of the literatures focusing on the following areas: experts search systems, experts profiling and ontology in the experts search system. The development of the framework involves extraction processes from web resources, for example, resume, individual websites, organization websites and research centers websites, to build a glossary of word index that characterizes experts. This glossary index becomes the basis for ontology development. In this research, the ontology development integrates the use of thesaurus, Association of Computing Machinery (ACM) taxonomy and ICT domain experts. The ontology that had enriched the glossary index provides the semantics ability in the ICT expert directory. There are three components in the framework; they are extraction of data, development of ontology, and validation of ICT expert ontology by domain experts. The development of this framework can guide the development of concept-based ontology that is relevant and accurate to user's needs.

Efficient Scene Change Detection and Camera Motion Annotation for Video Classification

In modern era, Internet plays a key role in accessing and fetching web information and web resources from World Wide Web (WWW). The websites act as a medium for retrieving information from the web. Although it increases the data retrieval and users interactions, it also opens the gate for various types of attacks. For example, spams in the websites attract various Internet users. It has been observed from the literature that many authors attempted to detect the web spam using various machine learning techniques. However, none of these techniques used deep learning architecture for detection of hidden patterns. Hence, in this paper, a deep learning algorithm, i.e., Recurrent Neural Networks (RNN), has been used for the classification of spam nodes. We devise here a framework called FS <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> RNN: Feature Selection Scheme using Recurrent Neural Networks. In this framework, the dataset is preprocessed before applying RNN in which principal component analysis (PCA) is used for dimension reduction on the data set and recursive feature elimination (RFE) is used for feature selection. The accuracy of the proposed framework, when compared before and after preprocessing, is improved by 24.2 %, which is excellent result.