PLAN RADA SEMINARA ZA NOVEMBAR 2019. GODINE
UTORAK, 05.11.2019 u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Božidar (Bob) Travica, University of Manitoba, Canada
A BIG DATA VIEW OF THE CUSTOMER
Leading companies model the customer differently than it is usually done. The change is triggered by the emergence of Big Data. Owing to the expansion of e-commerce carried out by the globalizing Internet and mobile technologies, the customer has become global. This customer is signified by the online behavior that consists of searching the Internet, moving in the cyber and physical worlds, online shopping and payment, and casual or purposive action in the social media space. This online behavior is tracked by new information technologies for storing and processing data characterized by a big variety of types, volume, and velocity of creation, that is, Big Data. The presentation explains modeling of the customer with Big Data and its implications for managerial and professional decision making in terms of benefits and risks. Guidelines for risk management are proposed.
UTORAK, 12.11.2019. u 14:15, sala 301f, MI SANU:
Nemanja Milenković, Fakultet organizacionih nauka, Beograd
METODOLOGIJA OTKRIVANJA NESTANDARDNIH OPSERVACIJA U K–DIMENZIONOM PROSTORU
Predmet istraživanja je formiranje metodologije za otkrivanje multivarijacionih nestandardnih opservacija kroz unapređenje metode Ivanovićevog odstojanja. Pod pojmom „nestandardna opservacija” se podrazumeva ona opservacija koja je na neki način nekonzistentna sa preostalima iz posmatranog skupa. Otkrivanje multivarijacionih nestandardnih opservacija se najčešće sprovodi korišćenjem metode Mahalanobisovog odstojanja. Ivanovićevo odstojanje se koristi u cilju merenja intenziteta neke pojave, korišćenjem većeg broja izabranih indikatora. Unapređena metoda Ivanovićevog odstojanja testira značajnost svakog od posmatranih indikatora korišćenjem odgovarajuće F statistike. Kroz upotrebu definisanih procedura za eliminaciju i/ili selekciju indikatora, nova metoda teži formiranju „optimalnog” skupa indikatora, redukujući dimenziju posmatranog kompleksnog problema.
UTORAK, 19.11.2019. u 14:15, sala 301f, MI SANU:
Aleksa Radosavčević, Branko Arsić, Marko Stojanović, Lazar Krstić, PMF Kragujevac i Aleksandar Glišović, ComTrade Digital Services
DEEP VORTEX TEAM - IBM DATA CHALLENGE WINNERS
On September 7th, IBM and Institute for Contemporary Sciences organized a datathon - data science competition for start-ups, professionals, academics and students. The competition was held in ICT Hub in Belgrade, where thirteen teams participated and met dataset and task requirements for the first time. The main goal was to develop a system for determining financial stability and capacity to sell distributed goods, for clients of large logistic company - Nelt. Proposed solution should automate Nelt's decision-making process regarding optimal credit line level which should be offered to each client. As the largest distribution and logistics company in the region, Nelt has a portfolio of a very diverse customer base, which differ in size, needs, procedures, geographies and policies. Sometimes clients unintentionally end up with more products than they could manage, or with liquidity issues and debt problems, all adding up to potential losses for Nelt. In order to help Nelt to maintain stable and long term relations with all of them, proposed solution should establish equilibrium between different needs of internal and external stakeholders: finance department with the goal to maintain company's financial stability, sales department to sell more goods and co-operant resellers to receive optimal amount of goods. The winner of the grand prize was Deep Vortex team from Kragujevac. This time they will present their solution in more detail manner and present all the challenges they were faced with during the competition, bearing in mind they just had 24 hours to design, test and present methodology of proposed solution.
UTORAK, 26.11.2019. u 14:15, sala 301f, MI SANU:
Kristina Kostić, Matematički fakultet, Beograd
REŠAVANJE PROBLEMA p-MEDIJANE SA RAVNOTEŽNIM OGRANIČENJIMA PRIMENOM HIBRIDNE MATEHEURISTIČKE METODE
Lokacijski problemi predstavljaju značajnu klasu problema optimizacije sa širokom primenom u praksi, posebno pri optimizaciji troškova u transportnim i telekomunikacionim mrežama. Generalno, lokacijski problemi podrazumevaju određivanje optimalnih lokacija za izgradnju objekata koji pružaju neku vrstu usluge, kao i pridruživanje svakog korisnika jednom ili više uspostavljenih uslužnih centara sa ciljem minimizacije ili maksimizacije određene funkcije cilja pri uslovima koji zavise od karakteristika konkretne mreže. Problem $p$-medijane (engl. p-median problem), jedan od značajnih pripadnika ove klase problema, podrazumeva lokaciju p uslužnih objekata i alokaciju korisnika uspostavljenim objektima sa ciljem minimizacije transportnih troškova. U okviru predavanja biće prezentovana varijanta problema p-medijane koja uključuje dodatna ograničenja sa ciljem balansiranja procesa opsluživanja korisnika (engl. p-median problem with balance constraints, p-MBC}). Kako p-MBC spada u klasu NP-teških problema kombinatorne optimizacije, razmatran je metaheuristički pristup rešavanju ovog problema. Preciznije, predložena je hibridizacija dveju metaheurističkih metoda: pohlepne stohastičko-adaptivne procedure pretrage (GRASP) i osnovne metode promenljivih okolina (BVNS). Elementi obe metode prilagođeni su karakteristikama p-MBC problema. Hibridna GRASP-BVNS metoda je testirana na poznatim instancama iz literature, kao i na novom skupu generisanih instanci. Dobijeni rezultati su upoređeni sa rezultatima egzaktnog rešavača.
RUKOVODIOCI SEMINARA
MI SANU
Vera Kovačević-Vujčić
Milan Dražić
FON
Zorica Bogdanovic
Marijana Despotovic-Zrakic
IEEE
Bozidar Radenkovic
