Back to Search Start Over

Analiza urinarnih biomarkera u svrhu predviđanja raka gušterače

Authors :
Ban, Idora
Domazet-Lošo, Mirjana
Publication Year :
2023
Publisher :
Sveučilište u Zagrebu. Fakultet elektrotehnike i računarstva., 2023.

Abstract

Karcinom gušterače izrazito je smrtonosna vrsta raka čijoj niskoj stopi preživljavanja najviše pridonosi činjenica da se simptomi počinju pojavljivati tek kada je bolest u uznapredovalom stadiju. U okviru ovoga rada proučavan je javno dostupni skup podataka koji uključuje vrijednosti urinarnih biomarkera kod zdravih osoba i osoba oboljelih od raka gušterače. Odabrane su i primijenjene metode strojnog učenja koje bi na temelju navedenog skupa mogle predvidjeti rak gušterače ili omogućiti dijagnosticiranje u ranijem stupnju bolesti. Najbolji rezultati dobiveni su korištenjem algoritama slučajna šuma, gradient boosting i stroj potpornih vektora. Implementirana je binarna i višeklasna klasifikacija, a evaluirani modeli uspoređivani su međusobno i s već postojećim analizama. Pancreatic cancer is an extremely deadly type of cancer whose low survival rate is attributable to the fact that symptoms only begin to appear when the disease has reached an advanced stage. In this thesis, a publicly available data set was studied, consisting of the values of urinary biomarkers in healthy people and patients with pancreatic cancer. Multiple machine learning methods were chosen and applied, assuming that they can use the aforementioned data set to predict pancreatic cancer or make earlier detection possible. The best results were obtained by using random forest, gradient boosting and support vector machine algorithms. Both binary and multiclass classification was implemented and the evaluated models were compared with each other and with existing analyses.

Details

Language :
Croatian
Database :
OpenAIRE
Accession number :
edsair.od......4131..8eefb8c64124a49f73d67fc44847263a