Naslov
Forest Covertype Prediction based on cartographic parameters using neural network

Autori
Lazar Dašić

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), ostalo

Izvornik
Ri-STEM-2021 Proceedings / Lorencin, Ivan ; Baressi Šegota, Sandi ; Car, Zlatan - , 2021, 141-145

ISBN
978-953-8246-22-7

Skup
International Student Scientific Conference (Ri-STEM 2021)

Mjesto i datum
Rijeka, Hrvatska, 10.06.2021. - 11.06.2021

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Artificial intelligence, Forest cover type, Multiclass classification, Sparse categorical crossentropy

Sažetak
The study shows success of Artificial Neural Network in making right prediction of forest cover type of certain area based on that area cartographic parameters. The study evaluated area of Roosevelt National Forest in Colorado. Artificial neural network used for making predictions is Multilayer perceptron using Stochastic Gradient Descent as an optimizer and Sparse categorical crossentropy as loss function. The result of the study showed that right neural network model can make over 90% correct predictions, which vastly transcend traditional statistical models. This success means that similar model can be used for making predictions in other areas around the world, which can decrease costs of managing natural resource inventory and even make managing possible for inaccessible areas.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Računarstvo, Temeljne tehničke znanosti

POVEZANOST RADA