Mikroservisna web aplikacija za usporedbu ML algoritama za predviđanje bankrota

master's thesis

• Title (english): Microservice web application for comparing ML algorithms for bankruptcy prediction
• Author: Sven Blagaić
• Mentor: Marina Bagić Babac (mentor)
• Committee member: Marina Bagić Babac (predsjednik povjerenstva)
• Committee member: Mirjana Domazet-Lošo (član povjerenstva)
• Committee member: Marko Horvat (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-10, Croatia

Abstract

Zadnjih godina povećao se broj bankrota raznih tvrtki zbog različitih ekonomskih i neekonomskih razloga. Predviđanje bankrota pokušava detektirati ovakve slučajeve temeljem raznih financijskih značajki kako bi se spriječilo njihovo buduće pojavljivanje. Cilj ovog rada bio je razviti web aplikaciju čija je svrha omogućiti lakšu usporedbu algoritama strojnog učenja koji se koriste za predviđanje bankrota. Dodatno, napravljena je detaljna usporedba korištenih algoritama koristeći rezultate dobivene iz aplikacije. Pet modela koji su se koristili bili su stablo odluke, slučajna šuma, k najbližih susjeda, neuronska mreža - višeslojni perceptron i stroj potpornih vektora te su bili testirani na četiri različite mjere vrednovanja - točnost, preciznost, odziv i F1 mjera. Odziv, odnosno udio stvarno pozitivnih primjera, je smatran najbitnijom mjerom jer propuštanje uočavanja slučaja bankrota ima veće posljedice od pogrešnog klasificiranja solventne tvrtke kao bankrotirajuće.

Abstract (english)

Company bankruptcy cases have risen in recent years due to different economic and non-economic factors. Bankruptcy prediction attempts to detect these events based on different financial features in order to prevent future occurences. The aim of this thesis was to develop a web application whose purpose is to enable easier comparison of different machine learning algorithms used for bankruptcy prediction. In addition, the algorithms were compared in detail using the results gained from the application. The five models in question were Decision tree, Random forest, K-nearest neighbors, Neural network - multilayer perceptron and Support vector machine, and they were tested on four different metrics – accuracy, precision, recall and F1 score while Type I and Type II errors were also considered. Recall, or the true positive rate, was regarded as the most important metric since failing to identify a bankruptcy case is usually more costly than falsely interpreting a solvent company as bankrupt.

• Keywords: Strojno učenje; Stablo odluke; Slučajna šuma; K najbližih susjeda; Neuronska mreža; Stroj potpornih vektora
• Keywords (english): Machine Learning; Decision Tree; Random Forest; K-nearest Neighbors; Neural Network; Support Vector Machine
• Language: english
• URN:NBN: urn:nbn:hr:168:748022
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 08:26:40