Identifikacija in analiza oglasnih člankov

PIČULIN, KRISTJAN (author), Sadikov, Aleksander (mentor)

Za diplomsko nalogo sem ustvaril program, ki prepoznava ali je spletni članek oglas ali pravi članek in tudi analiziral rezultate, ki jih je program dal. Analiziral sem, zakaj program določene članke označi kot prave članke ali oglase, zakaj nekatere članke narobe označi in kaj najbolj vpliva na njegovo delovanje. Zanimalo me je predvsem, kako ločiti navadne članke od oglasov. Program sem ustvaril v programskem jeziku python. Uporabil sem knjižnice selenium, pyqt, sklearn in podobne. Dosegel sem kar dobro in uspešno delovanje programa glede na moje cilje ter odkril mnogo zanimivega glede oglasov in novic.

• Language: Slovenian
• Keywords: strojno učenje, oglasi, oglaševanje, SVM, Naivni Bayes, SGD, TF-IDF
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2018
• PID: 20.500.12556/RUL-103532
• Publication date in RUL: 19.09.2018

Secondary language

• Language: English
• Title: Discovery and analysis of advertisements from textual data

Abstract

For my thesis i made a program, that recognizes if a web article is a pai advertisement or if it is a real news article and also analized the results that were made by the program. I analized why articles are classified the way they are, why are some articles misclassified and what things affect how program is recognizing articles. I was especially interested in a way to separate news articles and advertisements. The program was made in Python programming language. I used libraries such as: pyqt, sklearn and similar. I was quite successful in making the program work the way i wanted and i also found out many interesting things about articles and advertisements.

• Keywords: machine learning, paid articles, advertisement, SVM, Naive Bayes, SGD, TF-IDF