Seminar for
DECISION MAKING – THEORY, TECHNOLOGY AND PRACTICE

 

PROGRAM

---

Plan rada Seminara Odlučivanje - teorija, tehnologija, praksa za SEPTEMBAR 2023.

---

ČETVRTAK, 07.09.2023. u 13:00, Pariske Komune bb, Niš i Live stream Niš
Sanja Stevanović, Mathematical Institute of the Serbian Academy of Sciences and Arts
PREDICTING THE SHAPE OF LOADS FOR AN OFFICE CELL WITH AN OVERHANG FROM A SMALL NUMBER OF BUILDING ENERGY SIMULATIONS
Horizontal, rectangular window overhang is a simple, yet effective shading device. We study here the shapes of heating, cooling and lighting loads, as well as equivalent primary energy needs, in an office cell model with a single window as a function of the overhang depth and height, for a large set of 729,000 building energy simulations performed with EnergyPlus for the office cell model under various climates, obstacles, window orientations and heating and cooling set points. Simulation results indicate that the heating and cooling loads are rather smooth functions with a relatively slowly changing gradient, while the lighting load is more serrated, with significant jumps occurring along specific depth-height lines when the overhang depth becomes too large. We further measure the effectiveness of artificial neural networks and gradient-boosted tree ensembles (XGBoost) in approximating complete sets of simulated heating, cooling and lighting loads and equivalent primary energy from relatively small samples. Training of three artificial neural network models with differing depths and three XGBoost models with differing learning rates for each load and each combination of climate, obstacles, orientation and heating and cooling set points shows that XGBoost models are more precise and obtained more quickly than neural networks trained with pytorch. Experiences from this study are implemented in an openly available Python library, whose aim is to quickly construct surrogate models for these loads that depend on the depth and height of the overhang for the selected window in an arbitrary building model.

---

ČETVRTAK, 14.09.2023. u 13:00, Pariske Komune bb, Niš i Live stream Niš
Vladimir Simić, University of Belgrade, Faculty of Transport and Traffic Engineering
DECISION-MAKING FOR EMERGING SUSTAINABLE TRANSPORTATION MODELS
The lecture named “Decision-Making for Emerging Sustainable Transportation Models” aims to cover two important aspects:

1. Sustainable Urban Parcel Delivery; and
2. Sustainable Mobility Sharing Systems.

The Agenda for Problem 1 is the following:

• Transportation;
• Problem Definition & Research Questions;
• Urban Parcel Delivery;
• Definition of Urban Parcel Delivery Strategies;
• Definition of Evaluation Criteria;
• The Results in the Context of the City of Belgrade; and
• Findings & Implications & Major Conclusions.

The Agenda for Problem 2 is the following:

• Smart cities & Global Initiatives;
• Transportation & Global Initiatives;
• Concept of Shared Mobility;
• “Green” System of Shared Mobility;
• Problem Definition;
• Possible Implementation Barriers;
• Definition of Sustainable Mobility Sharing Systems;
• Definition of Evaluation Criteria;
• The Results in the Context of the City of Belgrade; and
• Findings & Implications & Major Conclusions.

---

ČETVRTAK, 21.09.2023. u 13:00, Live stream Niš
Željko Stević, University of East Sarajevo, Faculty of Transport and Traffic Engineering Doboj
INFLUENCE OF SENSITIVITY ANALYSIS ON DECISION-MAKING: EXAMPLES FROM LOGISTICS AND TRAFFIC
Sensitivity analysis represents changing criteria weights in the process of decision-making. It is an almost indispensable part of checking decision-making today and has a very large influence on the final ranking of alternatives. Regarding this, some examples of logistics and traffic are shown to demonstrate the influence of changing criteria weights. Various scenarios have been formed in which the significance of criteria has been reduced in intervals of 5-95 percent. Results show that sensitivity analysis has an influence on decision-making, even in some cases has a large influence.

---

ČETVRTAK, 28.09.2023. u 13:00, Live stream Niš
Anđelka Zečević, Mathematical Institute of the Serbian Academy of Sciences and Arts
MAPPING OF DISEASE NAMES TO DISEASE CODES BASED ON NATURAL LANGUAGE PROCESSING TECHNIQUES
Information aggregation from various gen, disease, and gen-disease databases into a unique database would enable researchers to analyze and compare valuable domain findings more conveniently and systematically. However, the aggregation poses numerous challenges due to non-uniform information annotation across the databases. One such challenge is mapping a disease name into a standardized disease code (DOID). This talk will present the benefits and limitations of using Natural Language Processing text representation techniques such as BioNLP off-the-shelf embeddings and the PubMedBERT language model and share some qualitative results related to the DisGeNET database.

---

Predavanja su namenjena sirokom krugu slusalaca, ukljucujuci studente redovnih i doktorskih studija. Seminar će se održavati svakog drugog četvrtka od 13:00 - 14:00^h, CIITLAB, Elektronski fakultet Niš, Aleksandra Medvedeva 14, Niš

---