Jyrki Nummenmaa

Constructing OLAP cubes based on queries

An On-Line Analytical Processing (OLAP) user often follows a train of thought, posing a sequence of related queries against the data warehouse. Although their details are not known in advance, the general form of those queries is apparent beforehand. Thus, the user can outline the relevant portion of the data posing generalised queries against a cube representing the data warehouse.Since existing OLAP design methods are not suitable for non-professionals, we present a technique that automates cube design given the data warehouse, functional dependency information, and sample OLAP queries expressed in the general form. The method constructs complete but minimal cubes with low risks related to sparsity and incorrect aggregations. After the user has given queries, the system will suggest a cube design. The user can accept it or improve it by giving more queries. The method is also suitable for improving existing cubes using respective real MDX queries.

Ontologies with Semantic Web/Grid in Data Integration for OLAP

Traditionally, data used in OLAP (online analytical processing) have been limited to the contents of the data warehouse of a company. However, the needs for analysis are often more demanding and data are needed from different sources. In this article, we study how the semantics of data sources can be described to allow combining data from several sources into an OLAP cube. We apply Semantic Web technologies for defining an OWL/RDF ontology for OLAP data sources and OLAP cubes. These definitions are then utilised in OLAP cube formation by posing an OWL/RDF ontology-based query against them. We use Grid technologies to enhance the efficiency of processing and ensuring security. Our primary interest is in the cube construction (i.e., ETL process), and we assume that standard OLAP methods can be used for the actual analysis. Our tests show that the proposed approach can speed up the construction of an OLAP cube for ad hoc queries by supporting a high-level query language and reducing the amount of required data.

Normalising OLAP cubes for controlling sparsity

A poorly designed OLAP (on-line analytical processing) cube can have a size much larger than the volume of information, potentially leading to problems with performance and usability. We give a new normal form for OLAP cube design and synthesis and decomposition algorithms to produce normalised OLAP cube schemata. OLAP cube normalisation controls the structural sparsity resulting from inter-dimensional functional dependencies. We assume that functional dependencies are used to describe the constraints of the application universe of discourse. Our methods help the user to identify cube schemata with structural sparsity, and to change the design in order to obtain more economy of space.

Towards Lightweight Requirements Documentation

Most requirements management processes and associated tools are designed for document-driven software development and are unlikely to be adopted for the needs of an agile software development team. We discuss how and what can make the traditional requirements documentation a lightweight process, and suitable for user requirements elicitation and analysis. We propose a reference model for requirements analysis and documentation and suggest what kind of requirements management tools are needed to support an agile software process. The approach and the reference model are demonstrated in Vixtory, a tool for requirements lightweight documentation in agile web application development.

Constructing compact rectilinear planar layouts using canonical representation of planar graphs

We present a new linear-time algorithm to construct a rectilinear planar layout (horvert-representation, visibility representation) for a given planar graph. Our approach is based on the canonical representation of planar graphs and it is basically different from previous algorithms. If we direct the edges from lower-numbered vertices to higher-numbered vertices and there are n vertices out of which k vertices have out-degree greater than in-degree, then the maximum width of the constructed layout is. ∑i=1nmax{dout(vi),0}⩽2n−4−(k−2)⩽2n−4 , and the maximum height is h + 1 ⩽ n, where h is length of the longest directed path. We discuss the selection of a good canonical numbering to be used when constructing layouts. We also show how our algorithm can be applied to compute planar layouts for planar graphs using other drawings for vertices than horizontal segments. In these layouts the drawings for vertices may have arbitrary nonequal sizes and shapes.

Mining Itemset-based Distinguishing Sequential Patterns with Gap Constraint

Mining contrast sequential patterns, which are sequential patterns that characterize a given sequence class and distinguish that class from another given sequence class, has a wide range of applications including medical informatics, computational finance and consumer behavior analysis. In previous studies on contrast sequential pattern mining, each element in a sequence is a single item or symbol. This paper considers a more general case where each element in a sequence is a set of items. The associated contrast sequential patterns will be called itemset-based distinguishing sequential patterns (itemset-DSP). After discussing the challenges on mining itemset-DSP, we present iDSP-Miner, a mining method with various pruning techniques, for mining itemset-DSPs that satisfy given support and gap constraint. In this study, we also propose a concise border-like representation (with exclusive bounds) for sets of similar itemset-DSPs and use that representation to improve efficiency of our proposed algorithm. Our empirical study using both real data and synthetic data demonstrates that iDSP-Miner is effective and efficient.

Model-driven approach to developing domain functional requirements in software product lines

Existing product requirements form a rich source for domain requirements analysis in software product lines (SPLs). Most existing domain analysis techniques depend on domain experts’ experience and manual operation to identify the commonalities and variabilities of product requirements. They often demand a high level of manual effort and a large up-front investment, which can present a prohibitive barrier for SPL adoption. This study proposes a model-driven approach to semi-automatically derive domain functional requirements (DFRs) from product functional requirements (PFRs). Based on the linguistic characterisation of a domains action-oriented concerns, the authors apply Fillmores semantic framework to functional requirements and define metamodels for PFRs and DFRs. Functional requirements of existing products are constructed as corresponding PFR models. Following the proposed merging and refinement rules, the authors approach automates the transformation from PFR models into DFR models by merging the same or similar PFRs and analysing their commonality and variability. The resulting DFR models can serve as an initial basis of the SPL. The authors demonstrate the authors approach using an example of a home security system (HSS) SPL and give a preliminary evaluation. The authors approach provides a rigorous model-based support for DFRs development and complements existing domain analysis techniques with less time and effort.

Using the Entity-Attribute-Value Model for OLAP Cube Construction

When utilising multidimensional OLAP (On-Line Analytic Processing) analysis models in Business Intelligence analysis, it is common that the users need to add new, unanticipated dimensions to the OLAP cube. In a conventional implementation, this would imply frequent re-designs of the cube’s dimensions. We present an alternative method for the addition of new dimensions. Interestingly, the same design method can also be used to import EAV (Entity-Attribute-Value) tables into a cube. EAV tables have earlier been used to represent extremely sparse data in applications such as biomedical databases. Though space-efficient, EAV-representation can be awkward to query.

Functional Dependencies in Controlling Sparsity of OLAP Cubes

We will study how relational dependency information can be applied to OLAP cube design. We use dependency information to control sparsity, since functional dependencies between dimensions clearly increase sparsity. Our method helps the user in finding dimensions and hierarchies, identifying sparsity risks, and finally changing the design in order to get a more suitable result. Sparse raw data, a large amount of pre-calculated aggregations, and many dimensions may expand the need of the storage space so rapidly that the problem cannot be solved by increasing the capacity of the system. We give two methods to construct suitable OLAP cubes. In the synthesis method, attributes are divided into equivalence classes according to dependencies in which they participate. Each equivalence class may form a dimension. The decomposition method is applied when candidates for dimensions exist. We decompose dimensions based on conflicts, and construct new cubes for removed dimensions until no conflicts between dimensions exist.

A generic data model with a decomposition operation for DRM interoperability

The digital rights management (DRM) systems attempt to protect digital content, which is by nature vulnerable to unauthorized distribution and use. The incompatibility of various DRM systems, however, hampers the diffusion of DRM. To increase DRM interoperability, we propose a generic rights model which illustrates the constructs and the relationships of rights. The model supports a decomposition operation, which allows the separation of different logical parts of rights. The decomposition provides a basis to improve interoperability between DRM systems and to maximize rights exporting.