Lecture
The information needed by the decision maker (DM) is rarely present in pure form, and useful information has to be extracted from a large amount of data. This process, as well as the corresponding line of research, has come to be called Data Mining, which consists of two concepts: searching for valuable information in a large database (Data) and mining ore (Mining). Both require either sifting through an enormous amount of raw material, or intelligent exploration and search for the sought-after values. The process of Data Mining proceeds at three levels, which is reflected in the translation of its name into Russian using the three concepts discussed above — as the mining of data, the extraction of information, and the acquisition of knowledge (see Fig. 4.2).
Unlike traditional artificial intelligence systems, the intelligent data search and analysis technology Data Mining, which combines the mining of data, the extraction of information, and the acquisition of knowledge, does not model natural intelligence but enhances its capabilities with the power of modern computers, search systems, and data warehouses.
DMs derive maximum benefit from information when it is accurate, complete, and knowledge can easily be extracted from it. In practice, however, information is usually noisy, and the amplitude of the useful signal is often comparable to the amplitudes of side effects, which can lead to erroneous assessments and decisions. In addition, information from structured repositories can be combined with information from unstructured sources, with access to it provided to various groups of users with different expectations regarding the ways it is presented. Ultimately, DMs need information and knowledge that correspond to their unique business processes and serve as a guide for decision-making at their own level. Among the information systems that satisfy the listed requirements, the following groups are distinguished :
Please note!
Classes of tasks that the specified information systems make it possible to solve. Data analytics — the calculation of specified indicators and statistical characteristics of an organization's activity based on retrospective information from databases.
Information visualization — a clear graphical and tabular representation of the available information.
Knowledge extraction — the determination of relationships and interdependencies among business processes based on existing information. This class can include tasks such as testing statistical hypotheses, clustering, and finding associations and temporal patterns.
Simulation analysis — conducting computational experiments with formalized (mathematical) models describing the behavior of complex systems over a certain period of time, in order to analyze the possible consequences of making managerial decisions.
Control synthesis — the determination of the set of admissible control actions ensuring the achievement of a given goal.
Control optimization — selecting, from the set of admissible controls, those that provide the most effective (with respect to the chosen criterion) progress toward the stated goal.
Comments