Pattern-based aggregate (PBA) queries constitute an important and widely used type of analytical queries in sequence OLAP (S-OLAP) systems. Unfortunately, finding accurate answers to PBA queries in the S-OLAP system is often very expensive both in terms of time and memory consumption. In this paper we propose an efficient and easily maintainable index structure called the ABBA Index, which addresses the problem of PBA query processing. Experiments conducted using the KDD Cup data and public transport passengers' travel behavior data show that our index outperforms state-of-the art solutions while requiring much less memory. The ABBA Index can be easily extended to support pattern-based aggregate queries over hierarchy (PBA-H), a novel class of analytical queries which we introduce as the second main contribution of the paper. Sensitivity, scalability and complexity analysis of the ABBA Index is also provided.

Managing tens of thousands of Database-as-a-Service instances presents significant challenges in fine-tuning their configurations to optimize database workload performance. Since individually tuning each instance is impractical, DBAs typically rely on universal, one-for-all configuration templates, which often fail to meet the specific requirements of certain instances. In this paper we explore methods that DBAs could use to automatically divide the Databaseas-a-Service instances into a small, manageable number of clusters with similar performance-to-configuration profiles. In this way, each group could share a single configuration template to maximize performance of its instances. We investigate whether such Database-as-a-Service instance clustering can be based solely on workload query frequencies. We validate our approach through an extensive set of experiments.

Data binarization involves converting a continuous data attribute into a finite set of binary attributes while minimizing information loss. It plays a crucial role in feature engineering in the data mining analysis. Data binarization simplifies data, improves model training quality, enhances model performance and interpretability of results, helping in understanding complex patterns. In this paper we present an original data binarization framework, called angle-based data binarization, that converts continuous attributes into discrete binary attributes. The proposed framework allows not only to simplify machine learning models, but can also lead to the improvement of the accuracy of a number of well-known traditional machine learning methods. We present results of an extensive series of experiments which evaluate the efficiency of the proposed method in the area of data classification. Using popular classification algorithms, we compared classification quality achieved on source datasets with classification quality achieved on their binarized versions. We also discuss binary attribute pruning, based on elimination of attributes with poor discriminative power.

Database Intrusion Detection (DID) is critical to systems that do not fully apply the Principle of Least Privilege and users are able to use their privileges to compromise data privacy. Existing DID tools focus on analyzing SQL texts and they miss the cases of query equivalence and do not reveal data sources hidden behind views, synonyms, aliases. We propose to use Query Execution Plans (QEPs) to detect misuse queries and we describe a QEP-based DID System.

Recently, numerous machine learning (ML) techniques have been applied to address database performance management problems, including cardinality estimation, cost modeling, optimal join order prediction, hint generation, etc. In this paper, we focus on query optimizer hints employed by users in their queries in order to mask some Query Optimizer deficiencies. We treat the query optimizer hints, bound to previous queries, as significant additional query metadata and learn to automatically predict which new queries will pose similar performance challenges and should therefore also be supported by query optimizer hints. To validate our approach, we have performed a number of experiments using real-life SQL workloads and we achieved promising results.

Many algorithms are designed to analyse volumes of data automatically in an efficient way so that the users don't have to look through that massive amount of data manually for generating various association rules among them. Apriori algorithm, which is the most famous and frequently used data mining algorithm. Our main focus is to parallelize the Apriori algorithm in such a new way that when we will implement on a large database , it will lead to less time consuming and fast execution for generating frequent itemset.

Nowadays everyone can review everything. Online customer-opinion platforms often help potential buyers take a decision. Sometimes, however, the multitude of contradictory opinions may confuse customers. Submitting a review requires time and effort, yet it only benefits others. Therefore jokes and shill reviews represent quite a percentage of reviews, because an average reviewer has no motivation to submit reviews. Reviews with Revenue in Reputation (RRR) Method is designed to encourage reviewers by gamification. RRR allows customers to easily spot credible reviews and restricts the number of reviews an unreliable reviewer can submit.

Anomaly detection in versatile financial data streams is a vital business problem. Existing IT solutions for business anomaly detection usually rely on explicit Complex Event Processing or near-real time Business Activity Monitoring. In this paper we argue that business anomaly detection should be considered an implicit infrastructural BPM service and we propose a corresponding Solution Pattern. We describe how a Business Anomaly Detector can be architectured and designed in order to handle fast dynamic streams of business objects in BPM environments. The presented solution has been practically verified in Oracle SOA/BPM Suite environment which handled real-life financial controlling business processes.

It is a challenge to design a well balanced reputation system for an environment with millions of users. A reputation system must also represent user reputation as a value which is simple and easy to compare and will give users straightforward suggestions who to trust. Since reputation systems rely on feedbacks given by users, it is necessary to collect unbiased feedbacksIn this paper we present a controversial, yet innovative reputation system. Hidden and Indirect (Probabilistically Estimated) Reputations - HIPER Method splits user reputation into two related values: Hidden Reputation (HR) is directly calculated from a set of feedbacks, Indirect Reputation (IR) is a probabilistically estimated projection of the hidden reputation and its value is public. Such indirect connection between received feedbacks and a visible reputation value allows users to provide unbiased feedbacks without fear of retaliation.

Rapid growth of spatial datasets requires methods to find (semi-)automatically spatial knowledge from these sets. Spatial collocation patterns represent subsets of spatial features whose instances are frequently located together in a spatial neighborhood. In recent years, efficient methods for collocation discovery have been developed, however, none of them assume limited size of the operational memory or limited access to memory with short access times. Such restrictions are especially important in the context of the large size of the data structures required for efficient identification of collocation instances. In this work we present and compare three algorithms for collocation pattern mining in a limited memory environment. The first algorithm is based on the well-known joinless method introduced by Shekhar and Yoo. The second and third algorithms are inspired by a tree structure (iCPI-tree) presented by Wang et al. In our experimental evaluation, we have compared the efficiency of the algorithms, both on synthetic and real world datasets.

In spatial databases, Collocation Pattern Discovery is a very important data mining technique. It consists in searching for types of spatial objects that are frequently located together. Due to high requirements for CPU, memory or storage space, such data mining queries are often executed at times of low user activity. Multiple users or even the same user experimenting with different parameters can define many queries during the working hours that are executed, e.g., at off-peak night-time hours. Given a set of multiple spatial data mining queries, a data mining system may take advantage of potential overlapping of the queried datasets. In this paper we present a new method for concurrent processing of multiple spatial collocation pattern discovery queries. The aim of our new algorithm is to improve processing times by reducing the number of searches for neighboring objects, which is a crucial step for the identification of collocation patterns.

Internet is a great medium that allows us to share knowledge with people from all over the world. On the other hand, information from the internet is unreliable because one cannot tell whether it was provided by an expert or a novice. In many cases it does not matter because everybody has the right to his/her opinion about for example a video clip. However there are situations when the credibility of the information is vitally important. Often one wants to find a solution to an important problem, therefore she/he needs some method to ignore jokes, spam and wrong answers by unexperienced users and consider only credible answers. We have designed Asymptotic Trust Algorithm (ATA) to manage users’ reputation in customer-to-customer environment like auction sites. In this paper we present an adaptation of ATA which allows to managed reputation of users of internet forums. We belive that our method will be useful for every internet forum which cares about reliability.

Over the last several years, processing of high performance data streams has become very important in various domains. A new type of data processing is needed for applications where input data streams are modeled as multimedia data streams, such as audio and video feeds. For example, in the public safety sector, monitoring and automatic identification of particular individuals suspected of terrorist or criminal activity requires the processing of complex audio and video streams, which is beyond the capabilities of a typical data stream management system (DSMS). The concept of a multimedia data stream management system (MMDSMS) has recently been introduced in order to effectively process continuous queries over dynamic multimedia data streams. In this paper, we address MMDSMS functionalities related to speaker recognition problems in the area of detecting individuals who may pose security threats. We focus on audio feed processing using our novel similarity-based join and on parameterization of the multimedia signal for the process of recognition. We propose a set of signal parameters which a clearly discriminate among individual voices by describing the signal using a homomorphic processing method. Our research was primarily focused on assessing the applicability of cepstral analysis in speech recognition systems, based on a set of acquired digitized voice samples. We developed a research prototype to assess the proposed concepts, and verified the effectiveness of our framework in a lab environment.

Frequent itemset mining is often regarded as advanced querying where a user specifies the source dataset and pattern constraints using a given constraint model. In this chapter we address the problem of processing sets of frequent itemset queries, which brings the ideas of multiple-query optimization to the domain of data mining. The most attractive method of solving the problem with respect to possible practical applications is Common Counting which consists in concurrent execution of the queries using Apriori with the integration of scans of the parts of the database shared among the queries. The major advantage of Common Counting over its alternatives is its applicability to arbitrarily large batches of queries. If the memory structures of all the queries to be processed by Common Counting do not fit together in main memory, the set of queries has to be partitioned into subsets processed in several phases. We formalize the problem of dividing the set of queries for Common Counting as a specific case of hypergraph partitioning and provide a comprehensive overview of query set partitioning algorithms proposed so far.

We consider the problem of executing collocation pattern queries in limited memory environments. Our experiments show that if the memory size is not sufficient to hold all internal data structures used by the iCPI-tree algorithm, its performance decreases dramatically. We present a new method to efficiently process collocation pattern queries using materialized, improved candidate pattern instance tree. We have implemented and tested the aforementioned solution and shown that it can significantly improve the performance of the iCPI-tree algorithm.

Overload management has become very important in public safety systems that analyse high performance multimedia data streams, especially in the case of detection of terrorist and criminal dangers. Efficient overload management improves the accuracy of automatic identification of persons suspected of terrorist or criminal activity without requiring interaction with them. We argue that in order to improve the quality of multimedia data stream processing in the public safety arena, the innovative concept of a Multimedia Data Stream Management System (MMDSMS) using load-shedding techniques should be introduced into the infrastructure to monitor and optimize the execution of multimedia data stream queries. In this paper, we present a novel content-centered load shedding framework, based on searching and matching algorithms, for analysing video tuples arriving within multimedia data streams. The framework tracks and registers all symptoms of overload, and either prevents overload before it occurs, or minimizes its effects. We have extended our Continuous Query Language (CQL) syntax to enable this load shedding technique. The effectiveness of the framework has been verified using both artificial and real data video streams collected from monitoring devices.

Online auctions have become a big business and the number of auction site users is growing rapidly. These virtual mar- ketplaces give traders a lot of opportunities to find a contracting party. However, lack of physical contact between users decreases the degree of trust. Auction portals require an efficient mechanism for building trust between participants, whereas most of them provide simple participation counts for reputation rating. Moreover, a single opinion has virtually no effect on a big online store that already has many reputation points, so buyers are very hesitant to give negative feedback for fear of retaliation. Consequently, almost no negative feedback is provided 1 . In this paper we introduce a new trust system called Asymp- totic Trust Algorithm (ATA) which prevents many fraud attempts and still is both simple and easy to understand for most users. Our new method can be applied in addition to the participation counts systems currently used by Allegro, eBay and most of other online auction sites because it does not require any additional information other than positive, negative or neutral feedback on transactions. Most importantly, ATA encourages users to submit unbiased com- ments, regardless of the number of previous transactions.