Metoda razvrščanja z združevanjem najbližjih sosedov v programu Orange

Šteblaj, Jurij (author), Zupan, Blaž (mentor)

Metoda razvrščanja z združevanjem najbližjih sosedov gradi filogenetska drevesa iz matrike razdalj med objekti. Uporablja se v bioinformatiki za napovedovanje evolucijskih razmerij med biološkimi vrstami. Kljub uporabnosti na širšem področju aplikacij pa ta metoda le redkokdaj zaide med programe za odkrivanje znanj iz podatkov. V sklopu diplomske naloge smo zato to metodo implementirali v obliki gradnika v programu Orange. Razvili smo tudi nekaj metod vizualizacije zgrajenih filogenetskih dreves in te preizkusili na klasičnih podatkih s področja podatkovnega rudarjenja. Rezultati pričajo o uporabnosti metode izven področja bioinformatike.

• Language: Slovenian
• Keywords: strojno učenje, združevanje najbližjih sosedov, bioinformatika, Orange
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2017
• PID: 20.500.12556/RUL-94873
• Publication date in RUL: 08.09.2017

Secondary language

• Language: English
• Title: Clustering by Neighbour Joining and its Implementation in Orange

Abstract

Neighbour joining builds phylogenetic trees from distance matrices. It is mainly used in bioinformatics for inference of evolutional relationships between species and prediction of common ancestors. Despite its usefulness in various applications it is rarely available in data mining programs. For this reason we implemented neighbour joining as a widget in general-purpose data mining suite Orange. We also developed several methods for visualisation of the inferred phylogenetic trees. Using our implementation on several use cases we have demonstrated that neighbour joining and constructed clustering trees are useful in data mining tasks outside the scope of bioinformatics.

• Keywords: machine learning, neighbour joining, bioinformatics, Orange