Integracija pripomočka za uravnoteževanje podatkov v paket Orange

LEMUT, MARK (author), Sadikov, Aleksander (mentor)

Statistične analize ali statistični testi se lahko izvajajo nad neuravnoteženimi podatkovnimi zbirkami, čeprav so statistični testi najbolj zanesljivi na uravnoteženih podatkih. Programov za uravnoteževanje podatkovnih zbirk ni veliko. Cilj naloge je bil narediti algoritem za uravnoteževanje, ki ga je razvil Aleš Smodiš dostopen za širšo množico, ki delajo s podatki in se ne ukvarjajo s programiranjem. Zato se je algoritem implementiral v brezplačni, odprtokodni program Orange, v obliko widgeta oz. pripomočka. Orange je preprost za uporabo in nudi veliko opcij za analizo, vizualizacijo in delo s podatki. Ponuja možnost uporabe strojnega učenja za klasifikacijo in regresijo.

• Language: Slovenian
• Keywords: widget, Orange, uravnoteževanje podatkov, predpriprava podatkov
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2018
• PID: 20.500.12556/RUL-100514
• Publication date in RUL: 23.03.2018

Secondary language

• Language: English
• Title: Integrating a data balancing widget into Orange machine learning suite

Abstract

Statistical tests can sometimes be performed on unbalanced data sets, even though they are most reliably performed on balanced data sets. There are not a lot of programs that can balance a data set. The assignments goal was to make the algorithm that was developed by Aleš Smodiš more accessible for an audience, that works with data and does not know how to program. Because of that, the algorithm was implemented into a free open source program called Orange. Orange is a simple program and offersa lot of options for data analysis, visualization, and for working with data in general. It offers the use of machine learning for classification and regression.

• Keywords: widget, Orange, data balancing, data pre-preparation