Analytical data processing methods for decision support

For some time now, the modern level of development of hardware and software has made possible the widespread maintenance of databases of operational information at all levels of management. In the course of their activities, industrial enterprises, corporations, departmental structures, government and administrative bodies have accumulated large amounts of data. They contain great potential for extracting useful analytical information, on the basis of which it is possible to identify hidden trends, build a development strategy, and find new solutions.

In recent years, a number of new concepts of storage and analysis of corporate data have taken shape in the world:

Data warehouses1 (Data Warehouse) 

On-Line Analytical Processing (OLAP)

Data mining - IAD (Data Mining)

This article is devoted to an overview of these concepts, as well as to the proof of their complementarity in support of management decision-making.

1. Warehouses (warehouses) of data

In the field of information technology, two classes of systems have always coexisted [16, p. 49]:

systems focused on operational (transactional) data processing; in the English-language literature they are often referred to as OLTP (On-Line Transaction Processing), as opposed to OLAP - on-line analytical processing [55]; A. A. Sakharov [15, p. 55] defines them as "data processing systems" (SOD);

systems focused on analytical data processing - decision support systems (DSS), or Decision Support Systems (DSS).

At the first stages of informatization, it is always necessary to put things in order precisely in the processes of daily routine data processing, which is what traditional ODS are focused on, therefore, the advanced development of this class of systems is quite understandable.

Systems of the second class - DSS - are secondary in relation to them. A situation often arises when data in an organization accumulates with a number of unrelated ODS, largely duplicating each other, but not being consistent in any way. In this case, it is practically impossible to obtain reliable complex information, despite its apparent excess.

The goal of building a corporate data warehouse is to integrate, update and reconcile operational data from heterogeneous sources to form a single consistent view of the control object as a whole. At the same time, the concept of data warehouses is based on the recognition of the need to separate the datasets used for transactional processing and the datasets used in decision support systems. Such division is possible by integrating the detailed data disaggregated in ODS and external sources into a single repository, their coordination and, possibly, aggregation. W. Inmon, the author of the concept of data warehouses [42], defines such warehouses as: desktop support job description

"Subject-oriented,

integrated,

unchanging,

maintaining chronology

data sets organized to support management ”designed to act as“ the one and only source of truth ”providing managers and analysts with the reliable information they need for operational analysis and decision support.

The concept of data warehouses is not just a single logical view of the organization's data, but the actual implementation of a single integrated data source. An alternative way to form a unified view of corporate data in relation to this concept is to create a virtual source based on distributed databases of various ODS. Moreover, each query to such a source is dynamically translated into queries to the source databases, and the results obtained are on the fly agreed upon, linked, aggregated and returned to the user. However, despite the external elegance, this method has a number of significant disadvantages.