20.500.12556/DKUM-63737 Razvoj odločitvenega modela za izbiro multimodalne biometrične nadzorne tehnologije Development of Decision Making Model for Selection of Multi-modal Biometric Control Technology System V magistrskem delu se ukvarjamo z razvojem prototipnega sistema za podporo odločanju pri izbiri multimodalne biometrične nadzorne tehnologije. V teoretičnem delu so predstavljene najbolj razširjene metode biometrične identifikacije in verifikacije. Nato smo predstavili sisteme za strojno učenje z metodo podatkovnega rudarjenja. Natančno smo predstavili tudi programsko orodje Orange in utemeljili, zakaj smo ga uporabili. V empiričnem delu naloge smo razvili prototip večkriterijskega odločitvenega modela za pomoč svetovalcem/prodajnikom pri izbiri topologije sistema za potrebe varovanja določenega objekta. Model smo razvili s pomočjo sistema na podlagi strojnega učenja Orange. V modelu smo uporabili naslednje klasifikacijske metode: Naive Beyes, Neural Network, k NN, SVM, Random Forest in Classification Tree. Model smo naučili in validirali na dejanskih podatkih obravnavanega podjetja. Kot najtočnejša metoda pri klasifikaciji se je izkazala metoda klasifikacije Random Forest z natančnostjo klasifikacije CA 98,63 %. Razviti model je pokazal zadovoljivo natančnost pri ustrezni izbiri sistema biometrične identifikacije in verifikacije. In this master degree we are dealing with development of prototype model for decission making support when choosing multimodal biometric system for future installation. Furthermore the most common methods of biometric identification and verification are presented here. Then we have introduced most common analytical suites based on machine learning and data mining methods. We have also described data mining suite Orange in details and justified why we chose this suite. In the empiric part of this scientific work we developed a prototype decision support system to help advisers / sales management in the selection of the system's topology for the purposes of protecting a particular object. The model was developed with Orange suite. We used following classification methods in the model: Naive Beyes, Neural Network, k-NN, SVM, Random Forest in Classification Tree. The model has been learned and validated on actual data of the company observed. During the work, we find method Random Forest as the most accurate classification method in this research with accuracy of classification CA 98,63%. The developed model showed a satisfactory precision in choosing an appropriate system of biometric identification and verification. pristopna kontrola biometrija strojno učenje klasifikacija večkriterijski odločitveni modeli access control biometrics machine learning classification decision support model true true false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2016-09-15 11:53:23 2016-10-18 13:30:16 2022-07-03 03:06:41 2019-07-11 13:37:38 2016 0 Kranj 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 7801363 URN:SI:UM:DK:IH6KZUHH MAG_Bozic_Tomaz_2016.pdf MAG_Bozic_Tomaz_2016.pdf 1 63EB70D2786B67661F1F90D0E4CE7A75 9bde2d080d0c4077adee8d949f7fb0d25eff1a30e0cee580f33cfeb5b48d130f 7ab03ed0-7c0d-11eb-bb7a-00155d0001ca https://dk.um.si/Dokument.php?lang=slv&id=104612 Fakulteta za organizacijske vede 0 0 0