20.500.12556/RUL-119126 Analiza časa izdelave oglasov Analysis of advert production time Spletno oglaševanje predstavlja čedalje večji delež celotnega oglaševanja, s tem pa se razvijajo tudi različne platforme za izdelavo spletnih oglaševalskih kampanj. Naš cilj je bil ugotoviti, kateri atributi najbolj vplivajo na potreben čas za izdelave kampanje na podlagi podatkov, ki nam jih je priskrbelo podjetje Celtra d.o.o. Atribute ocenjujemo z RReliefF-om in Boruto, nato zgradimo več regresijskih napovednih modelov in najuspešnejšega med njimi razložimo z uporabo metode SHAP. Ugotovimo, da na čas izdelave najbolj vpliva število komponent oglasov v kampanji, število uporabnikov in to, koliko dela na kampanji opravijo grafični oblikovalci. Online advertising amounts to an increasingly big share of all advertising. Along with that, more and more different platforms for the production of online advertising campaigns are being developed. Our goal is to find out, which attributes most affect the necessary production time of online advertising campaigns. We use the data about campaigns production provided by Celtra d.o.o. We evaluate attributes with the use of RReliefF and Boruta, then we build several different machine learning models for campaign production time prediction. With the use of Shapley additive explanations, we explain the most successful of our models. We find out that the number of components used, the number of users, and the percentage of production time done by graphic designers have the greatest impact on the production time. naključni gozdovi regularizacija Tihonova metoda podpornih vektorjev odločitveno drevo RReliefF Boruta vrednosti Shapley Shapleyeve aditivne razlage pomembnost atributov oglaševanje random forests ridge regression support vector machine decision tree RReliefF Boruta Shapley values Shapley additive explanations feature importance advertising true false false Slovenski jezik Angleški jezik Diplomsko delo/naloga 2020-09-03 10:55:01 2020-09-03 10:55:06 2023-12-28 09:25:35 0000-00-00 00:00:00 2020 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 25783 27705091 Torkar_Miha_-_Analiza_casa_izdelave_oglasov.pdf Torkar_Miha_-_Analiza_casa_izdelave_oglasov.pdf 1 9A492D1DB6D4B5CC804337074EC25E17 8fa29c79476eb13e1c60bdc5809772c459866f13bf7659eecd9f03d13d521548 bed8c2e6-a1b9-11eb-a523-00155dcfd717 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=133810 Fakulteta za računalništvo in informatiko Fakulteta za matematiko in fiziko 0 0 0