Прегледај по Аутор "Janković, Radmila"

Сада се приказује 1 - 3 од 3
  • Сличица
    Ставка
    Analyzing the Еffects of Мobility and Season on COVID-19 Cases Using Negative Binomial Regression: a European Case Study
    (Faculty of Electrical Engineering, University of East Sarajevo, 2021) Janković, Radmila; Amelio, Alessia; Ćosović, Marijana
    This paper develops a Generalized Linear Model using the Negative Binomial Regression with log link function to analyze the effects of mobility trends and seasons on COVID-19 cases. The data of four European countries was used, namely Austria, Greece, Italy, and Czech Republic. The dataset includes daily observations of registered COVID-19 cases, and the data of six types of mobility trends: retail and recreation, grocery and pharmacy, parks, transit stations, workplaces, and residential mobility for the period Feb 15 - Nov 15, 2020. The results suggest that the number of COVID-19 cases differs between seasons and different mobility trends.
  • Сличица
    Ставка
    CNN Classification of the Cultural Heritage Images
    (Faculty of Electrical Engineering, University of East Sarajevo, 2020) Ćosović, Marijana; Janković, Radmila
    The cultural heritage image classification represents one of the most important tasks in the process of digitalization. In this paper, a deep learning neural network was applied in order to classify images of architectural heritage belonging to ten categories, in particular: (i) bell tower, (ii) stained glass, (iii) vault, (iv) column, (v) outer dome, (vi) altar, (vii) apse, (viii) inner dome, (ix) flying buttress, and (x) gargoyle. The Convolutional neural network was used for image classification, with the same architecture applied on two sets of the data: the full dataset consisting of 10 categories as well as dataset with 5 different image categories. The results show that both architectures performed well and obtained accuracy of up to 90%.
  • Сличица
    Ставка
    Time Series Prediction of Air Pollutants
    (Faculty of Electrical Engineering, University of East Sarajevo, 2019) Janković, Radmila; Ćosović, Marijana; Amelio, Alessia
    Pollution levels are highly dependent on the meteorological parameters, as the weather conditions dictate pollution dispersion and concentration. With the rise of global environmental protection initiatives, there is also a need for accurate prediction of pollution levels. This paper presents a time series prediction of NO2 and CO given four meteorological parameters: (i) air pressure, (ii) relative humidity, (iii) average daily temperature, and (iv) wind speed, using a Nonlinear Autoregressive Exogenous (NARX) neural network. The research is a case study of three European countries: (i) Serbia, (ii) Bosnia and Herzegovina, and (iii) Italy, and involves data from 2014 to 2016 for a total of 1096 instances. The results show that the best prediction accuracy is obtained for CO for data regarding Italy and Bosnia and Herzegovina, and for NO2 for data regarding Serbia. Moreover, the best predictor variables of NO2 are air pressure and relative humidity, followed by the wind speed. The best predictor variables of CO are pressure and temperature for Bosnia and Italy, and wind speed for Serbia