Refinement of a graph convolutional neural network approach applied to classify cancer types

Entity: Universitat Rovira i Virgili (URV)

Confidenciality: No

Education area(s): Enginyeria de la Seguretat Informàtica i Intel·ligència Artificial

APS: NO

Title in different languages: Refinement of a graph convolutional neural network approach applied to classify cancer types

Abstract: The objective of the master’s thesis is to study and refine the model architecture of a previously presented approach for improving the prediction accuracy in each cancer types. This work has four GCNN models which are PPI, PPIS, COEX and COEXS, using data from TCGA of cancer and normal tissue as the input depending on the gene expression. After refinement, the prediction accuracy of four GCNN models achieved the outstanding 89-96% among 34 classes and reduced predictive errors in original model of cancer classes that present problems.

Subject: Càncer

Academic year: 2021-2022

Language: en

Work's public defense date: 2022-02-09

Subject areas: Computer engineering

Student: Ousuwan, Aphanchanok

Department: Enginyeria Informàtica i Matemàtiques

Creation date in repository: 2022-11-25

Keywords: graph convolutional neural networks, cancer classification, deep learning

Title in original language: Refinement of a graph convolutional neural network approach applied to classify cancer types

Access Rights: info:eu-repo/semantics/openAccess