Most financial institutions mainly use numerical statistics to detect anomalous (malpractice) activity. The textual content in financial records however contains precious information which to date has not been effectively used for detection of anomalous behaviors by users because these are often unintelligible, cluttered with abbreviations, numbers and symbols, which makes it difficult to build a framework system that can coherently understand and draw conclusions. Rule-based techniques have been proposed but such systems are easy to elude, as they are difficult to generalize and do not scale up. The work presented in this paper differs from previous work in that we exclusively base anomalous activities on text (excluding numerical values) in financial records and treat this as a classification problem for a deep learning network. We propose four solutions using deep learning techniques on textual data to distinguish between normal with anomalous behaviors of the users. The results of our experiments convincingly show that use of the textual content in financial records yields greater accuracy in anomalous behavior detection. They also suggest that deep learning is a viable and effective solution for real time anomaly detection by financial institutions.

We study the impact of preprocessing techniques on performance and effectiveness of the similarity measures for uncertain time series. Some existing work on uncertain time series use the same similarity measures developed for standard time series, to which we refer as traditional similarity measures. More recently, a number of new similarity measures have been proposed for uncertain time series, to which we refer as uncertain similarity measures. However, they have been shown not to be as effective as the traditional measures. In this work, we show that the performance of uncertain similarity measures can be improved through preprocessing techniques. We establish this through extensive experiments using the UCR benchmark data. Our results in fact indicate that the uncertain similarity measures together with preprocessing outperform the traditional similarity measures.

Uncertain time series analysis is important in applications such as wireless sensor networks and location-based services. This has been the subject of some recent studies, and a number of solution techniques have been proposed for similarity search problems. We classify the proposed similarity measures into deterministic, which returns a value, and probabilistic, which returns a random variable. By means of our classification, we present an overview of the proposed similarity measures and evaluate them experimentally. We conducted a comprehensive performance evaluation of these techniques through numerous experiments using the well-known real-life UCR benchmark data. As the computational complexity of some of these similarity measures was very high, we devised an effective sampling-based heuristic method to complete the experiments which could not be done before. The results of our experimental evaluation and comparison provide useful insights and guidelines for researchers and practitioners in similarity search and analysis of uncertain time series data.

Indexing techniques are crucial for efficiency and scalability of processing queries over big data. Hive is a batch-oriented big data management engine that is well suited for data OLAP and data analysis applications. For very “selective” queries whose output sizes are a small fraction of the contributing data, the brute-force approach suffers from poor performance due to redundant disk I/O's or initiations of extra map operations. We make a first attempt and propose an index-based join technique to speed up the process and integrate it in Hive by mapping our design to the conceptual optimization flow. To evaluate the performance, we create and evaluate test queries on datasets generated using TPC-H benchmark. Our results indicate significant performance gain over relatively large data and/or highly selective queries having a two-way join and a single join condition.

Numerous real-life applications, such as wireless sensor networks and location-based services, generate large amount of uncertain time series, where the exact value at each timestamp is unavailable or unknown. In this paper, we formalize the notion of correlation for uncertain time series data and consider a family of probabilistic, threshold-based correlation queries over such data. The proposed formulation extends the notion of correlation developed for standard, certain time series. We show that uncertain correlation is a random variable approaching normal distribution. We also formalize the notion of uncertain time series normalization which is at the core of our correlation query processing approach, while it proves to be an important pre-processing technique in particular for pattern discovery tasks. The results of our numerous experiments indicate that, unlike in the standard time series, there is a trade-off between false alarms and hit ratios, which can be controlled by the probability threshold provided by users. Our results also offer users a guideline for choosing proper threshold values.

Fuzzy clustering is a popular method for modeling web usage data, and a number of techniques have been proposed. Performance of such techniques has been demonstrated through experiments using datasets which are often limited in the size and/or variety. This is mainly due to the difficulty in acquiring large real data, and also to the huge amount of time and effort required in performing experiments. We investigate ways to ensure dependability of such results and their analyses. For this we consider three issues. First we need to ensure that the clustering quality indices used for comparing different techniques are not biased towards any parameter specific to any of them. Second, more ground truth is provided by measuring the quality through an application of the usage model than through the clustering quality index alone. Third, given the limited data sets and experiments, use of statistical significance testing can provide more confidence in that the results obtained are not by mere chance. We present our approach for dependable performance analysis using some well-known fuzzy clustering techniques along with prediction quality used as the application specific metric.

There has been increasing interest for efficient techniques for fast correlation analysis of time series data in different application domains. We present three algorithms for (1) bivariate correlation queries, (2) multivariate correlation queries, and (3) correlation queries based on a new correlation measure we introduce using dynamic time warping. To support these algorithms, we use a variant of the Compact Multi-Resolution Index (CMRI). In addition to conventional nearest neighbor and range queries supported by CMRI, the proposed algorithms compute all answers to user-defined, ad hoc and parametric correlation queries. The results of our experiments indicate a speed-up of two orders of magnitude over the brute force algorithm, and an order of magnitude improvement on average, while offering more functionalities than provided by existing techniques such as StatStream and the Spatial Cone Tree.

Rank order correlation has been used extensively when the data is non-parametric or when the relationship between two variables is nonlinear and monotonic. In such cases, linear correlation measures, such as the product-moment coefficient, are inadequate and fail to detect correlative relations. We present a polyhedral indexing technique for rank order correlation queries for time series data. We use an interesting geometry interpretation of rank order correlation which lends itself to indexing by spatial indexes such as R-trees. Our experimental results indicate one to two orders of magnitudes improvement over sequential scan - the only alternative solution.

Heterogeneity of schema and data in information integration complicates metadata management and query processing. We consider a mediator-based approach (MI) to information integration and propose a model theoretic approach to describe integration. In this model, we view the schema of the integration as a 3D space, and assume it is a complete lattice. The first dimension in the model specifies the concepts (e.g., entity sets, relations, classes, etc), the second dimension indicates the data model in which a concept is represented (e.g., relational, semi-structured, object-oriented, etc), and the third dimension gives the application domain. We also introduce three basic transformations, called X-transform, Y-transform, and Z-transform, to all of which we refer as primitive queries and show how user queries can be expressed using primitive queries. We show a typical architecture and illustrate usefulness of this model as it generalizes the mediator based information integration in which the global schema is basically the least upper bound of all the points in the 3D space.

We present a novel approach to visualize Web usage patterns by closely coupling the visual rendering process to the data mining technique. In the first step we use relational fuzzy subtractive clustering as the mining technique to perform fuzzy clustering on Web usage sessions. In the second step, we use conventional metric multidimensional scaling to obtain an initial positional configuration in 3D space for the cluster centers, and then apply a modified Sammon mapping technique to further optimize the 3D positions. In the last step, we use the dominant membership values to assign positions to all the other sessions in the given dataset. This is computationally very efficient and at the same time retains the fidelity of the interrelationships much better. We have developed a running prototype of the proposed approach and have demonstrated the utility through experiments using several datasets, including a fairly large Web usage dataset of about 100,000 log records

We study suitable indexing techniques to support efficient exact match search in large biological sequence databases. We propose a suffix tree (ST) representation, called STA-DF, as an alternative to the array representation of ST (STA) proposed in [7] and utilized in [18]. To study the performance of STA and STA-DF, we develop a memory efficient ST-based Exact Match (STEM) search algorithm. We implemented STEM and both representations of ST and conducted extensive experiments. Our results indicate that the STA and STA-DF representations are very similar in construction time, storage utilization, and search time using STEM. In terms of the access patterns by STEM, our results show that compared to STA, the STA-DF representation exhibits better spatial and sequential locality of reference. This suggests that STA-DF would require less number of disk I/Os, and hence is more amenable to efficient and scalable disk-based computation.

A number of approaches which use model-based collaborative filtering (CF) for scalability in building recommendation systems in Web personalization have poor accuracy due to the fact that Web usage data is often sparse and noisy. Clustering, mining association rules, and sequence pattern discovery have been used to determine the access behavior model. Making use of some of the characteristics of the modeling process can provide significant improvements to recommendation effectiveness. In an earlier work, we introduced a fuzzy hybrid CF technique which inherits the advantages of both memory-based and model-based CF. In this paper, using relational fuzzy subtractive clustering as the first level modeling and then mining association rules within individual clusters, we propose a two level model-based technique, which is scalable and is an enhancement over association rule based recommender systems. Our results from comprehensive experiments using a large real life Web usage data and performance comparisons with memory-based and model-based approaches help substantiate this claim.

The OWL ontology explorer ONTOXPL is based on the web server tomcat. Standard HTML browsers can be used to interact with ONTOXPL. It is intended to complement existing ontology editors and does not offer any editing support. ONTOXPL uses the OWL DL reasoner RACER via its extensive query interface in order to support the intelligent exploration of OWL ontologies.

Numerous frameworks have been proposed in recent years for deductive databases with uncertainty. On the basis of how uncertainty is associated with the facts and rules in a program, we classify these frameworks into implication-based (IB) and annotation-based (AB) frameworks. We take the IB approach and propose a generic framework, called the parametric framework, as a unifying umbrella for IB frameworks. We develop the declarative, fixpoint, and proof-theoretic semantics of programs in our framework and show their equivalence. Using the framework as a basis, we then study the query optimization problem of containment of conjunctive queries in this framework and establish necessary and sufficient conditions for containment for several classes of parametric conjunctive queries. Our results yield tools for use in the query optimization for large classes of query programs in IB deductive databases with uncertainty.

Introduction Database system technology has reached a stage now in which there is a proliferation of independent systems storing and manipulating enormous amount of data. Unfortunately, these systems typically have their own data models, communication processing protocols, query processing systems, concurrency control protocols, consistency management, and other similar aspects of database systems. There is also an increasing need for Interoperability among these systems. Though considerable amount of research has been done in the area of database interoperability, most of it has resulted in solutions that are ad-hoc and procedural. We have developed a declarative environment in which multiple heterogeneous databases interoperate by sharing, interpreting, and manipulating information, in a uniform way. An important criterion for Interacting with multiple databases is the ability to query them in a manner independent of the discr

Article On implementing SchemaLog—a database programming language Share on Authors: Alanoly J. Andrews Concordia University, Montreal, Canada Concordia University, Montreal, CanadaView Profile , Nematollaah Shiri Concordia University, Montreal, Canada Concordia University, Montreal, CanadaView Profile , Laks V. S. Lakshmanan Concordia University, Montreal, Canada Concordia University, Montreal, CanadaView Profile , Iyer N. Subramanian Concordia University, Montreal, Canada Concordia University, Montreal, CanadaView Profile Authors Info & Claims CIKM '96: Proceedings of the fifth international conference on Information and knowledge managementNovember 1996 Pages 309–316https://doi.org/10.1145/238355.238565Online:12 November 1996Publication History 4citation247DownloadsMetricsTotal Citations4Total Downloads247Last 12 Months6Last 6 weeks1 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access