Modele şi metode robuste de prelucrare a datelor cu aplicaţii în ştiinţele naturii
Robust models and methods for data processing with applications in natural sciences
Obiectivul principal al prezentului proiect este dezvoltarea de metode si modele robuste de analiza si procesarea datelor cu impact in domeniul stiintelor naturii. Avem in vedere problematicile urmatoare: metode robuste de analiza datelor si metode de analiza multicriteriala pentru analiza datelor in stiintele naturii. Vom urmari solutiile bazate pe soft computing, cu preferinta metodele fuzzy. De exemplu, pls fuzzy, pcr fuzzy, biclustering fuzzy. Soft computing in general, metodele bazate pe multimi fuzzy in particular, s-au dovedit foarte eficiente in analiza datelor de origine chimica, fizica, biologica si geologica, datorita in special faptului ca datele din aceste domenii prezinta particularitati capturate cu succes de teoria multimilor fuzzy. Metode robuste de data mining cu aplicatii in diagnoza medicala. Metode robuste de data mining in contextul asigurarii anonimatului datelor, lucru extrem de important in majoritatea experimentelor din domeniul stiintelor naturii, stiintelor medicale si stiintelor psihologice. Modelarea formala a sistemelor soft bazate pe componente. Dezvoltarea unei metode robuste de specificare si constructie a modelului unui sistem, metoda cat mai abordabila, fara o investitie mare de timp si de efort intelectual din partea modelatorului. Dezvoltarea unui instrument de analiza a modelului creat, instrument care sa ofere rezultate despre comportamentul sistemului considerat. Dezvoltarea de metode si modele robuste pentru studierea metricilor soft utilizare in proiectarea orientata obiect a sistemelor soft. Cresterea programelor in dimensiune si complexitate a condus la costuri ridicate de dezvoltare si productivitate scazuta. Sistemele au devenit astfel inflexibile (nu pot fi adaugate cu usurinta functionalitati noi), monolitice (nu exista o functionalitate a sistemului bazata pe componente) si greu de intretinut (incercarea de a aduce noi modicari se soldeaza cu un lant nesfarsit de ajustari in multiple locuri).
The main objective of this research project is the development of robust methods and models of data analysis and data processing with impact in the field of natural sciences. We are considering the following problematics: robust methods and multicriterial analysis methods for data analysis in natural sciences. We are interested, especially, in methods based on the soft computing approach, preferably methods based on fuzzy sets, as, for example, fuzzy pls, fuzzy pcr, fuzzy biclustering. Soft computing in general, methods based on fuzzy sets in particular, proved to be very effective in analysing of data of chemical, physical, biological and geological origin, especially due to the fact that the data from these fields show particularities successfully captured by the fuzzy sets theory. Robust methods of data mining for medical diagnosis. Robust methods of data mining in the context of data anonimity maintenance, issue of extreme importance for the experiments in the domain of natural sciences, medical sciences and phychological sciences. Formal modeling of components based software systems. Development of a robust specification method and construction of the model of a system, a more approachable method, without a great initial investment of time and intellectual effort from the modeler. Development of an analysis tool of the created model, tool witch will offer result about the systems behaviour. Development of robust methods and models for the study of software metrics used in the object oriented systems design. The increase of programs in size and complexity lead to higher development costs and lower productivity. Software systems have such become inflexible (new functionality cannot be easily added), monolythical (the system does not provide a functionaly based on components) and difficult to maintain (the effort to bring new modifications leads to an unending chain of adjustments in multiple places).
© Prof.dr. Horia F. Pop