Iskanje in razvrščanje spletnih trgovin

BIRSA, ARON (author), Robnik Šikonja, Marko (mentor)

Cilj diplomske naloge je razvoj orodja, ki omogoča avtomatsko zaznavanje spletnih trgovin glede na tip izdelkov, ki jih ponuja. Spletne strani smo klasificirali v sedem vnaprej določenih kategorij: starine in zbirke, oblačila, zabavna elektronika, pohištvo, dom in vrt, nakit in pisarniški izdelke. Glavni problem je bil pridobivanje ustreznih podatkov za izgradnjo učne in testne množice ter klasificiranje spletnih strani. Uporabili smo naslednje metode strojnega učenja: naivni Bayesov klasifikator, k-najbližjih sosedov, metodo naključnih gozdov, nevronsko mrežo in metodo podpornih vektorjev. Najbolj obetavne rezulate smo dobili z metodo podpornih vektorjev.

• Language: Slovenian
• Keywords: specializirani iskalnik, podatkovno rudarjenje, strojno učenje, spletne trgovine, analiza besedil, naivni Bayesov klasifikator, k-najbližjih sosedov, metoda naključnih gozdov, nevronska mreža, metoda podpornih vektorjev.
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2017
• PID: 20.500.12556/RUL-88879
• Publication date in RUL: 24.01.2017

Secondary language

• Language: English
• Title: Search and classification of web shops

Abstract

The aim of the thesis was to develop a tool for automatic classification of online stores depending on the type of products they offer. Websites are classified into seven predefined categories: antiques and collectibles, cloth- ing, consumer electronics, furniture, home and garden, jewelry and office products. The main problem was getting relevant data to build a learning and test data set and classifying web sites. The following machine learning methods were used: naive Bayesian classifier, k-nearest neighbors algorithm, random forests, neural networks and support vector machine. The most promising result were obtained using the support vector machine classifier.

• Keywords: specialized search engine, data mining, machine learning, e- commerce, text analysis, naive Bayesian classifier, k-nearest neighbors algo- rithm, random forests, neural networks, support vector machine.