Refine
Year of publication
Document Type
- Article (26)
- Conference Proceeding (26)
- Report (9)
- Part of a Book (8)
- Doctoral Thesis (1)
- Preprint (1)
Keywords
- Skyline (3)
- Cardiology and Cardiovascular Medicine (2)
- Clinical Neurology (2)
- Advanced and Specialised Nursing (1)
- Advanced and Specialized Nursing (1)
- Borda Social Choice (1)
- Clustering (1)
- Datenbank (1)
- General Medicine (1)
- Multi-core (1)
Institute
- Fakultät für Angewandte Informatik (65)
- Institut für Informatik (65)
- Lehrstuhl für Datenbanken und Informationssysteme (64)
- Medizinische Fakultät (6)
- Universitätsklinikum (5)
- Lehrstuhl für Neurologie (4)
- Institut für Mathematik (1)
- Lehrstuhl für Psychiatrie und Psychotherapie (1)
- Mathematisch-Naturwissenschaftlich-Technische Fakultät (1)
- Professur für Experimentelle Allergologie und immunologische Intoleranz (1)
Skyline evaluation techniques (also known as Pareto preference queries) follow a common paradigm that eliminates data elements by finding other elements in the data set that dominate them. To date already a variety of sophisticated skyline evaluation techniques are known, hence skylines are considered a well researched area. Nevertheless, in this paper we come up with interesting new aspects. Our first contribution proposes so-called semi-skylines as a novel building stone towards efficient algorithms. Semi-skylines can be computed very fast by a new Staircase algorithm. Semi-skylines have a number of interesting and diverse applications, so they can be used for constructing a very fast 2-dimensional skyline algorithm. We also show how they can be used effectively for algebraic optimization of preference queries having a mixture of hard constraints and soft preference conditions. Our second contribution concerns so-called skyline snippets, representing some fraction of a full skyline. For very large skylines, in particular for higher dimensions, knowing only a snippet is often considered as sufficient. We propose a novel approach for efficient skyline snippet computation without using any index structure, by employing our above 2-d skyline algorithm. All our efficiency claims are supported by a series of performance benchmarks. In summary, semi-skylines and skyline snippets can yield significant performance advantages over existing techniques.
Complex application domains like outdoor activity platforms demand a powerful search interface that can adapt to personal user preferences and to changing contexts like weather conditions. Today most platforms offer a search technology known as Faceted Search, also named Parametric Search, where a user iteratively adapts his/her search parameters by a tedious and time-consuming trial-and-error process until the quality and quantity of the query results somehow corresponds to his/her expectations. This process gets even more cumbersome in mobile environments. Here we present a sophisticated approach called Preference Search, which we have prototypically implemented in a commercial outdoor activity platform. Preference Search replaces lengthy user sessions by one single user request. Technically, this request is automatically compiled into one single Preference SQL query, which efficiently retrieves those items that best match the user's expectations within the current context. A benchmark was applied to Faceted Search as well as Preference Search. The evaluation of the benchmark indicates that Preference Search substantially improves the user's search satisfaction in comparison to Faceted Search.
Preference queries become more and more important in applications like OLAP, data warehousing, or decision support systems. In these environments the Preference SQL GROUPING operation and aggregate functions are extensively used in formulating queries. In this report we present the full specification of the GROUPING operation in Preference SQL. This specification describes the grouping and aggregation known from standard SQL as well as the grouping with substitutable values (SV) semantics to allow a flexible and powerful grouping functionality in comparison to standard SQL. Furthermore, we introduce novel algebraic transformation laws for grouped preference queries and numerical ranking which are one of the most intuitive and practical type of queries. We explain how Preference SQL can be modified to integrate these optimization laws into the existing rule-based query optimizer. Our study upon the well-known TPC-H benchmark dataset shows that significant performance gains can be achieved.
Preferences are an important natural concept in real life and are well-known in the database and artificial intelligence community. The integration of preference queries in database systems enables satisfying search results by delivering best matches, even when no object in a dataset fulfills all preferences perfectly. Skyline queries are the most prominent representatives of preferences queries. The target is to select those tuples from a dataset that are optimal with respect to a set of designated preference attributes. But users do not only think of finding the Pareto frontier, they often want to find the best objects concerning an explicit specified preference order. While preferences themselves often are defined as general strict partial orders, almost all algorithms are designed to evaluate Pareto preferences combining weak orders, i.e., Skylines. In this paper, we consider general strict partial orders and we present a method to evaluate such explicit preferences by embedding any strict partial order into a lattice. This enables preference evaluation with specialized lattice based algorithms.
Given a data set, a top-k Skyline query returns the k most interesting elements of the Skyline query based on some kind of user-defined preference. That means, sometimes not only the Pareto frontier is of interest, but also the stratum, the level, behind the Skyline to get exactly the top-k objects from a partially ordered set stratified into subsets of non-dominated tuples. In this paper, we extend the definition of top-k Skyline to form multi-level Skyline sets. Multi-level Skylines are a variant of top-k Skylines which do not stop after k tuples, but compute all Skyline levels. We present a parallel algorithm for multi-level Skyline computation on multi-core architectures and demonstrate through extensive experimentation on synthetic and real data sets that our algorithms can result in a significant performance advantage over existing techniques.
In Data Mining large and increasing sets of data are becoming more and more common. In order to avoid losing the overview on these data-sets, preference queries are a very popular method to reduce quantities of data to high relevant information. Together with clustering methods like k-means, confusing sets of objects can be constituted and presented clearer in order to get a better overview. In this report we present on the one hand the Pareto-dominance as a very suitable and promising approach to cluster objects over better-than relationships. In order to meet someones desires, one can tip the balance of the final results to the more favored dimension if no decision for allocating objects is possible. On the other hand we introduce based on the Pareto-dominance an advanced clustering approach exploiting the Borda Social Choice voting rule to manage distances of different domains by equally weights during the clustering process.
For a given set of Skyline points, we consider the problem of finding a minimal set of new points which dominates all given Skyline points. The placement of the new points is subject to a given feasibility condition. We propose an optimal and efficient algorithm for the 2-dimensional case. Regarding the generalized d-dimensional (d≥3) problem we present different efficient heuristics approximating the minimal set of dominators.
Skyline queries enable satisfying search results by delivering best matches, even if the filter criteria are conflictive. The result of a Skyline query consists of those objects for which there is no dominating object in the input data set. Algorithms for Skyline computation are often classified into generic and index-based approaches. While there are uncountable papers on the comparison on generic algorithms, there exists only a few publications on the effect of index-based Skyline computation. In this technical report, we evaluate the most recent index-based Skyline algorithms BBS, ZSky, and SkyMap in order to find out which algorithm performs best. We conducted comprehensive experiments on different data sets and present some really unexpected outcomes.
Skyline processing has received considerable attention in the last decade, in particular when filtering the most preferred objects from a multi-dimensional set on contradictory criteria. Most of the work on Skyline computation focus on conventional databases, but stream data analysis becomes a high relevant topic in various academic and business fields. Nowadays, an enormous number of applications require the analysis of time evolving data and therefore the study of continuous query processing has recently attracted the interest of researchers all over the world. In this paper, we propose a novel algorithm called SLS for evaluating Skyline queries over continuous settings, and empirically demonstrate the advantage of this algorithm on artificial and real data. Our algorithm continuously monitors the incoming data and therefore is able to maintain the Skyline incrementally. For this, SLS utilizes the lattice structure a Skyline query constructs and analyzes the Skyline in linear time.
Preference algebra, an extension of the algebra of database relations, is a well-studied field in the area of personalized databases. It allows modelling user wishes by preference terms; they represent strict partial orders telling which database objects the user prefers over other ones. There are a number of constructors that allow combining simple preferences into quite complex, nested ones. A preference term is then used as a database query, and the results are the maximal objects according to the order it denotes. Depending on the size of the database, this can be computationally expensive. For optimisation, preference queries and the corresponding terms are transformed using a number of algebraic laws. So far, the correctness proofs for such laws have been performed by hand and in a point-wise fashion. We enrich the standard theory of relational databases to an algebraic framework that allows completely point-free reasoning about complex preferences. This black-box view is amenable to a treatment in first-order logic and hence to fully automated proofs using off-the-shelf verification tools. We exemplify the use of the calculus with some non-trivial laws, notably concerning so-called preference prefilters which perform a preselection to speed up the computation of the maximal objects proper.
In this demo paper we present a recommender system, which exploits the Borda social choice voting rule for clustering recommendations in order to produce comprehensible results for a user. Considering existing clustering techniques like k-means, the overhead of normalizing and preparing the preferred user data is dropped. In our demo showcase we facilitate a comparison of our clustering approach to the well known k-means++ with traditional distance measures.
anaging large and confusing sets of increasing data is a well-known problem in Data Mining. Since compromises in many use cases like Recommender Systems or preference-based applications are becoming more and more usual, it is very useful to cluster sets of promising results in order to get an overview and present them properly. In this paper we present the Pareto-dominance as a very suitable and promising approach to cluster objects over better than relationships. In order to meet someones desires, one can tip the balance of the final results to the more favored dimension if no decision for allocating objects is possible.
In this paper we present our demo application which allows preference-based search for interesting information in a data stream. In contrast to existing stream analysis services, the application uses the attributes of the stream records in combination with soft conditions to achieve the best possible result for a query.
Automatic recommendations are very popular in E-commerce, online shopping platforms, video on-demand services, or music-streaming. However, recommender systems often suggest too many related items such that users are unable to cope with the huge amount of recommendations. In order to avoid losing the overview in recommendations, clustering algorithms like k-means are a very common approach to manage large and confusing sets of items. In this paper, we present a clustering technique, which exploits the Borda social choice voting rule for clustering recommendations in order to produce comprehensible results for a user. Our comprehensive benchmark evaluation and experiments regarding quality indicators show that our approach is competitive to k-means and confirms the high quality of our Borda clustering approach.
