| Popis: |
U ovom radu je ukratko objašnjen algoritam Gradient Boost i prikazana matematika na kojoj je baziran i koju je objasnio Jerome H. Freidman. Gradient Boost je pokazao velik uspjeh u širokom spektru primjene u stvarnom svijetu. Također su definirani pojmovi funkcija gubitka (loss function), learning rate i overfitting. Funkcija gubitka i learning rate kao jedne od najznačajnijih faktora u Gradient Boost, te overfitting kao jedan od najvećih problema Gradient Boosta. Također je objašnjen primjer klasifikacije baze podatka koristeći Gradient Boost „Titanic- Top 1% with Gradient Boost Clasifier“. Ovim primjerom se pokušava odrediti koji putnik će preživjeti i koji neće preživjeti potonuće Titanika. Primjer je uzet sa stranice Kaggle i napisan je u Python-u. Kroz objašnjenje primjera smo prošli kroz danu bazu podataka, prilagodili varijable (skupine podatka) našim potrebama, odredili suodnose varijabli, kako je odabran Gradinet Boost i prilagodili hiperparametre. In this pape Gradient Boost Algoritham is briefly explained and math by Jerome H. Freidman on witch it is based. Gradient Boost has shown great success in wide rage of applications. The concepts of loss function, learning rate and overfitting are also defined. The function of loss and learning rate are one of the most important factors in Gradient Boost, and overfitting is one of the biggest problems of Gradient Boost. There is also an explanation of an database classification using Gradiet Boost „Titanic- Top 1% with Gradient Boost Clasifier“. Attempt of this exaple is to determined wich passenger will survive and wich will not survive the sinking of the Titanic.This example was taken from th Kaggle page and it was writen in Python. Going through exaplme , we went over given datebase, adjusted the variables to our needs, determined the correlations of the varibale, how was Gradient Boost selesceted and adjusted the hyperparamethrs. |
