Bitcoin (BTC) shares many characteristics with traditional stocks, but it is much more volatile since the cryptocurrency market is unregulated. The high volatility makes BTC a very high risk-high reward investment and predictive analysis can be very useful to obtain good returns and minimize risk. Taking Cocco et al. [1] as the primary reference, this thesis tries to reproduce their findings by building two BTC price forecasting models, Long Short-Term Memory (LSTM) and Bayesian Neural Network (BNN), and finding that the Mean Absolute Percentage Error (MAPE) is lower for the initial BNN model in comparison to the initial LSTM model. In addition to forecasting the value of BTC, a metric called trend% is developed to gauge the models’ ability to capture the trend of how the price varies from one timestep to the next and used to compare the trend prediction performance. It is found that both initial models make random predictions for the trend. Improvements like removing the stochastic component from the data and forecasting returns as opposed to price values show that both models show comparable performance in terms of both MAPE and trend%. The thesis concludes by discussing the future work that can be done to potentially improve the above models. One of the possibilities mentioned is to use on-chain data from the BTC blockchain coupled with the real-world knowledge of BTC exchanges and feed this as input features to the models.

Deep neural networks have been shown to be vulnerable to adversarial attacks. Typical attack strategies alter authentic data subtly so as to obtain adversarial samples that resemble the original but otherwise would cause a network's misbehavior such as a high misclassification rate. Various attack approaches have been reported, with some showing state-of-the-art performance in attacking certain networks. In the meanwhile, many defense mechanisms have been proposed in the literature, some of which are quite effective for guarding against typical attacks. Yet, most of these attacks fail when the targeted network modifies its architecture or uses another set of parameters and vice versa. Moreover, the emerging of more advanced deep neural networks, such as generative adversarial networks (GANs), has made the situation more complicated and the game between the attack and defense is continuing. This dissertation aims at exploring the venerability of the deep neural networks by investigating the mechanisms behind the success/failure of the existing attack and defense approaches. Therefore, several deep learning-based approaches have been proposed to study the problem from different perspectives. First, I developed an adversarial attack approach by exploring the unlearned region of a typical deep neural network which is often over-parameterized. Second, I proposed an end-to-end learning framework to analyze the images generated by different GAN models. Third, I developed a defense mechanism that can secure the deep neural network against adversarial attacks with a defense layer consisting of a set of orthogonal kernels. Substantial experiments are conducted to unveil the potential factors that contribute to attack/defense effectiveness. This dissertation also concludes with a discussion of possible future works of achieving a robust deep neural network.

Social media has become a primary means of communication and a prominent source of information about day-to-day happenings in the contemporary world. The rise in the popularity of social media platforms in recent decades has empowered people with an unprecedented level of connectivity. Despite the benefits social media offers, it also comes with disadvantages. A significant downside to staying connected via social media is the susceptibility to falsified information or Fake News. Easy accessibility to social media and lack of truth verification tools favored the miscreants on online platforms to spread false propaganda at scale, ensuing chaos. The spread of misinformation on these platforms ultimately leads to mistrust and social unrest. Consequently, there is a need to counter the spread of misinformation which could otherwise have a detrimental impact on society. A notable example of such a case is the 2019 Covid pandemic misinformation spread, where coordinated misinformation campaigns misled the public on vaccination and health safety. The advancements in Natural Language Processing gave rise to sophisticated language generation models that can generate realistic-looking texts. Although the current Fake News generation process is manual, it is just a matter of time before this process gets automated at scale and generates Neural Fake News using language generation models like the Bidirectional Encoder Representations from Transformers (BERT) and the third generation Generative Pre-trained Transformer (GPT-3). Moreover, given that the current state of fact verification is manual, it calls for an urgent need to develop reliable automated detection tools to counter Neural Fake News generated at scale. Existing tools demonstrate state-of-the-art performance in detecting Neural Fake News but exhibit a black box behavior. Incorporating explainability into the Neural Fake News classification task will build trust and acceptance amongst different communities and decision-makers. Therefore, the current study proposes a new set of interpretable discriminatory features. These features capture statistical and stylistic idiosyncrasies, achieving an accuracy of 82% on Neural Fake News classification. Furthermore, this research investigates essential dependency relations contributing to the classification process. Lastly, the study concludes by providing directions for future research in building explainable tools for Neural Fake News detection.

Distributed self-assessments and reflections empower learners to take the lead on their knowledge gaining evaluation. Both provide essential elements for practice and self-regulation in learning settings. Nowadays, many sources for practice opportunities are made available to the learners, especially in the Computer Science (CS) and programming domain. They may choose to utilize these opportunities to self-assess their learning progress and practice their skill. My objective in this thesis is to understand to what extent self-assess process can impact novice programmers learning and what advanced learning technologies can I provide to enhance the learner’s outcome and the progress. In this dissertation, I conducted a series of studies to investigate learning analytics and students’ behaviors in working on self-assessments and reflection opportunities. To enable this objective, I designed a personalized learning platform named QuizIT that provides daily quizzes to support learners in the computer science domain. QuizIT adopts an Open Social Student Model (OSSM) that supports personalized learning and serves as a self-assessment system. It aims to ignite self-regulating behavior and engage students in the self-assessment and reflective procedure. I designed and integrated the personalized practice recommender to the platform to investigate the self-assessment process. I also evaluated the self-assessment behavioral trails as a predictor to the students’ performance. The statistical indicators suggested that the distributed reflections were associated with the learner's performance. I proceeded to address whether distributed reflections enable self-regulating behavior and lead to better learning in CS introductory courses. From the student interactions with the system, I found distinct behavioral patterns that showed early signs of the learners' performance trajectory. The utilization of the personalized recommender improved the student’s engagement and performance in the self-assessment procedure. When I focused on enhancing reflections impact during self-assessment sessions through weekly opportunities, the learners in the CS domain showed better self-regulating learning behavior when utilizing those opportunities. The weekly reflections provided by the learners were able to capture more reflective features than the daily opportunities. Overall, this dissertation demonstrates the effectiveness of the learning technologies, including adaptive recommender and reflection, to support novice programming learners and their self-assessing processes.

In some scenarios, true temporal ordering is required to identify the actions occurring in a video. Recently a new synthetic dataset named CATER, was introduced containing 3D objects like sphere, cone, cylinder etc. which undergo simple movements such as slide, pick & place etc. The task defined in the dataset is to identify compositional actions with temporal ordering. In this thesis, a rule-based system and a window-based technique are proposed to identify individual actions (atomic) and multiple actions with temporal ordering (composite) on the CATER dataset. The rule-based system proposed here is a heuristic algorithm that evaluates the magnitude and direction of object movement across frames to determine the atomic action temporal windows and uses these windows to predict the composite actions in the videos. The performance of the rule-based system is validated using the frame-level object coordinates provided in the dataset and it outperforms the performance of the baseline models on the CATER dataset. A window-based training technique is proposed for identifying composite actions in the videos. A pre-trained deep neural network (I3D model) is used as a base network for action recognition. During inference, non-overlapping windows are passed through the I3D network to obtain the atomic action predictions and the predictions are passed through a rule-based system to determine the composite actions. The approach outperforms the state-of-the-art composite action recognition models by 13.37% (mAP 66.47% vs. mAP 53.1%).

The need of effective forecasting models for multi-variate time series has been underlined by the integration of sensory technologies into essential applications such as building energy optimizations, flight monitoring, and health monitoring. To meet this requirement, time series prediction techniques have been expanded from uni-variate to multi-variate. However, due to the extended models’ poor ability to capture the intrinsic relationships among variates, naïve extensions of prediction approaches result in an unwanted rise in the cost of model learning and, more critically, a significant loss in model performance. While recurrent models like Long Short-Term Memory (LSTM) and Recurrent Neural Network Network (RNN) are designed to capture the temporal intricacies in data, their performance can soon deteriorate. First, I claim in this thesis that (a) by exploiting temporal alignments of variates to quantify the importance of the recorded variates in relation to a target variate, one can build a more accurate forecasting model. I also argue that (b) traditional time series similarity/distance functions, such as Dynamic Time Warping (DTW), which require that variates have similar absolute patterns are fundamentally ill-suited for this purpose, and that should instead quantify temporal correlation in terms of temporal alignments of key “events” impacting these series, rather than series similarity. Further, I propose that (c) while learning a temporal model with recurrence-based techniques (such as RNN and LSTM – even when leveraging attention strategies) is challenging and expensive, the better results can be obtained by coupling simpler CNNs with an adaptive variate selection strategy. Putting these together, I introduce a novel Selego framework for variate selection based on these arguments, and I experimentally evaluate the performance of the proposed approach on various forecasting models, such as LSTM, RNN, and CNN, for different top-X% percent variates and different forecasting time in the future (lead), on multiple real-world data sets. Experiments demonstrate that the proposed framework can reduce the number of recorded variates required to train predictive models by 90 - 98% while also increasing accuracy. Finally, I present a fault onset detection technique that leverages the precise baseline forecasting models trained using the Selego framework. The proposed, Selego-enabled Fault Detection Framework (FDF-Selego) has been experimentally evaluated within the context of detecting the onset of faults in the building Heating, Ventilation, and Air Conditioning (HVAC) system.

The proliferation of semantic data in the form of RDF (Resource Description Framework) triples demands an efficient, scalable, and distributed storage along with a highly available and fault-tolerant parallel processing strategy. There are three open issues with distributed RDF data management systems that are not well addressed altogether in existing work. First is the querying efficiency, second is that solutions are optimized for certain types of query patterns and don’t necessarily work well for all types, and third is concerned with reducing pre-processing cost. Therefore, the rapid growth of RDF data raises the need for an efficient partitioning strategy over distributed data management systems to improve SPARQL (SPARQL Protocol and RDF Query Language) query performance regardless of its pattern shape with minimized pre-processing overhead. In this context, the first contribution of this work is a distributed RDF data partitioning schema called 3CStore that extends the existing VP (Vertical Partitioning) approach by using a subset of triples from the VP tables based on different join correlations. This approach speeds up queries at the cost of additional pre-processing overhead. To solve this, a relational partitioning schema called VPExp was developed by splitting predicates based on explicit type information of objects. This approach gains a significant query performance only for the specific type of query where the object is bound to a value for a particular predicate. To get efficient query performance on a wide range of query patterns, an improved solution is proposed by extending the existing Property Table approach to Subset-Property Table and combined with the VP approach. Further investigation on distributed RDF processing and querying systems based on typical use cases led to a novel relational partitioning schema called PTP (Property Table Partitioning) that further partitions the whole Property Table into the number of unique properties to minimize query input size and join operations during query evaluation. Finally, an RDF data management system based on the SPARQL-over-SQL approach called S3QLRDF is developed that generates the optimal query execution plan using statistics of PTP tables to provide efficient SPARQL query processing on a distributed system.

Tensors are commonly used for representing multi-dimensional data, such as Web graphs, sensor streams, and social networks. As a consequence of the increase in the use of tensors, tensor decomposition operations began to form the basis for many data analysis and knowledge discovery tasks, from clustering, trend detection, anomaly detection to correlationanalysis [31, 38]. It is well known that Singular Value matrix Decomposition (SVD) [9] is used to extract latent semantics for matrix data. When apply SVD to tensors, which have more than two modes, it is tensor decomposition. The two most popular tensor decomposition algorithms are the Tucker [54] and the CP [19] decompositions. Intuitively, they both generalize SVD to tensors. However, one key problem with tensor decomposition is its computational complexity which may cause system bottleneck. Therefore, two phase block-centric CP tensor decomposition (2PCP) was proposed to partition the tensor into small sub-tensors, execute sub-tensor decomposition in parallel and combine the factors from each sub-tensor into final decomposition factors through iterative rerefinement process. Consequently, I proposed Sub-tensor Impact Graph (SIG) to account for inaccuracy propagation among sub-tensors and measure the impact of decomposition of sub-tensors on the other's decomposition, Based on SIG, I proposed several optimization strategies to optimize 2PCP's phase-2 refinement process. Furthermore, I applied SIG and optimization strategies for data focus, data evolution, and focus shifting in tensor analysis. Personalized Tensor Decomposition (PTD) is proposed to account for the users focus given the observations that in many applications, the user may have a focus of interest i.e., part of the data for which the user needs high accuracy and beyond this area focus, accuracy may not be as critical. PTD takes as input one or more areas of focus and performs the decomposition in such a way that, when reconstructed, the accuracy of the tensor is boosted for these areas of focus. A related challenge of data evolution in tensor analytics is incremental tensor decomposition since re-computation of the whole tensor decomposition with each update will cause high computational costs and incur large memory overheads. Especially for applications where data evolves over time and the tensor-based analysis results need to be continuouslymaintained. To avoid re-decomposition, I propose a two-phase block-incremental CP-based tensor decomposition technique, BICP, that efficiently and effectively maintains tensor decomposition results in the presence of dynamically evolving tensor data. I further extend the research focus on user focus shift. User focus may change over time as data is evolving along the time. Although PTD is efficient, re-computation for each user preference update can be the bottleneck for the system. Therefore I propose dynamic evolving user focus tensor decomposition which can smartly reuse the existing decomposition result to improve the efficiency of evolving user focus block decomposition.

In videos that contain actions performed unintentionally, agents do not achieve their desired goals. In such videos, it is challenging for computer vision systems to understand high-level concepts such as goal-directed behavior. On the other hand, from a very early age, humans are able to understand the relation between an agent and their ultimate goal even if the action gets disrupted or unintentional effects occur. Inculcating this ability in artificially intelligent agents would make them better social learners by not just learning from their own mistakes, i.e, reinforcement learning, but also learning from other's mistakes. For example, this could greatly reduce the search space for artificially intelligent agents for finding the correct action sequence when trying to achieve a new goal, since they would be able to learn from others what not to do as well as how/when actions result in undesired outcomes.To validate this ability of deep learning models to perform this task, the Weakly Augmented Oops (W-Oops) dataset is proposed, built upon the Oops dataset. W-Oops consists of 2,100 unintentional human action videos, with 44 goal-directed and 33 unintentional video-level activity labels collected through human annotations. Inspired by previous methods on tasks such as weakly supervised action localization which show promise for achieving good localization results without ground truth segment annotations, this paper proposes a weakly supervised algorithm for localizing the goal-directed as well as the unintentional temporal region of a video using only video-level labels. In particular, an attention mechanism based strategy is employed that predicts the temporal regions which contributes the most to a classification task, leveraging solely video-level labels. Meanwhile, our designed overlap regularization allows the model to focus on distinct portions of the video for inferring the goal-directed and unintentional activity, while guaranteeing their temporal ordering. Extensive quantitative experiments verify the validity of our localization method.

Many real-world problems, such as model- and data-driven computer simulation analysis, social and collaborative network analysis, brain data analysis, and so on, benefit from jointly modeling and analyzing the underlying patterns associated with complex, multi-relational data. Tensor decomposition is an ideal mathematical tool for this joint modeling, due to its simultaneous analysis of such multi-relational data, which is made possible by the data's multidimensional, array-based nature. A major challenge in tensor decomposition lies with its computational and space complexity, especially for dense datasets. While the process is comparatively faster for sparse tensors, decomposition is still a major bottleneck for many applications. The tensor decomposition process results in dense (hence, large) intermediate results, even when the input tensor is sparse (or small). Noise is another challenge for most data mining techniques, and many tensor decomposition schemes are sensitive to noisy datasets; this is an inevitable problem for real-world data, which can lead to false conclusions. In this dissertation, I develop innovative tensor decomposition algorithms for mining both sparse and dense multi-relational data in a noise-resistant way. I present novel, scalable, parallelizable tensor decomposition algorithms, specifically tuned to be effective for dense, noisy tensors, and which maintain the quality of the resulting analysis. Furthermore, I present results on multi-relational data applications focusing on model- and data-driven computer simulation analysis, as well as social network and web mining, which demonstrate the effectiveness of these tensor decompositions.