Metodologia projektowania problemowo zorientowanych baz danych do systemów wielostopniowego podejmowania decyzji.

Przedstawiono metodologię projektowania problemowo zorientowanych baz danych przeznaczonych do systemów wielostopniowego podejmowania decyzji. Zaprezentowano zastosowanie tej metodologii do projektowania konkretnych systemów baz danych: systemów przeznaczonych do wspomagania podejmowania decyzji w procesie wielostopniowego rozpoznawania obiektów i identyfikacji wielostopniowej. Skoncentrowano się na następujących zagadnieniach: na problemie wyboru struktur danych i opracowaniu algorytmów zarządzania pamięcią BD do wielostopniowego rozpoznawania obiektów na problemie wyboru modelu BD do identyfikacji wielostopniowej, problemie równoważności modeli problemowo zorientowanych baz danych do wielostopniowej identyfikacji i wielostopniowego rozpoznawania oraz na opracowaniu problemowo zorientowanych języków baz danych.

---

In the monograph, we consider the issues the designing problem-oriented databases supporting specific decision systems for multistage identification and pattern recognition. The methods of designing such database matching specific needs of application are proposed. We deal with the following design issues: the choice of data structures and the development of memory management algorithms for multistage identification, multistage pattern recognition and the development of problem - oriented data base languages. The following problems are presented: ˇ The method of the design of external and conceptual data base models by the definition of the so-called segments. ˇ The external and conceptual data base models ˇ The data base manipulation language. ˇ The algorithms for data base management system (algorithms for data base manipulation functions and algorithms for memory management) for multistage identification and multistage pattern recognition data bases. ˇ Data structure problem formulation. The experiments were conducted for different data structures where the storage occupancy as well as the execution times of identification and recognition algorithms accessing a data base were studied. Taking into account the results achieved the choice method was proposed to choose the structure of the relations for the data base as well as the way of the decomposition of the learning sequence. ˇ The development of the mapping of logical structures onto physical structures (the method of data base organization) by means of indexes for multistage identification for different data base models (hierarchical, network and relational). ˇ Learning sequence decomposition problem formulation. Two ways of decomposition of the learning sequence for centralized data base as well as three decomposition ways for distributed data base were developed. The decomposition schemes were compared in the context of storage and access time requirements. Some decomposition theorems are proved. ˇ The problem of the data base model chose for the multistage identification was formulated. The storage occupancy theorems were proved for hierarchical and network models. Three models, i.e. hierarchical, network and relational were compared taking into account the storage occupancy and access time. ˇ The equivalence condition for problem-oriented data base models was formulated. It was proved that hierarchical, network and relational models of data structures and operations in data bases for multistage identification and pattern recognitionare equivalent. Data base and knowledge base were compared; it was proposed to use relational data base techniques for knowledge representation. In this paper, two problem-oriented data bases, i.e. IDEN for multistage identification and REC for multistage recognition, are presented.