Geografska segmentacija uporabnikov za uporabo v oglaševanju : diplomsko delo

Dolenc, Blaž (author), Zupan, Blaž (mentor)

V današnjem spletnem oglaševanju ni več edini cilj prikazati oglasa čim večjemu številu potencialnih kupcev, temveč si oglaševalci vse bolj prizadevajo oglas prikazati tistemu, ki ga bo najverjetneje zanimal. Na primer, če poznamo uporabnikovo okvirno lokacijo, lahko na podlagi prejšnjih obiskovalcev napovemo klik oglasa. Potrebo po geografski segmentaciji uporabnikov so zaznali tudi pri podjetju Zemanta, kjer so študentom zastavili izziv, pri katerem je bilo potrebno obiskovalce spletnih strani razdeliti glede na poštno številko iz katere prihajajo, ter to uporabiti kot podlago za napoved klika. Cilj naloge je bilo poiskati čim bolj smiselne skupine uporabnikov, ter jih ustrezno predstaviti, v drugem delu pa zgraditi napovedni model za napovedovanje klika na oglas, ki bo dosegal točnost napovedi AUC okoli 0,75. V nalogi poročamo o naši rešitvi tega problema, ki uporablja vrsto tehnik s področja strojnega učenja. Končna razdelitev uporabnikov, ki jo predlagamo, je obsegala 20 skupin, ki so se med seboj močno razlikovale glede na gostoto poselitve, urbanizacije in ostalih demografskih dejavnikov. Prikaz skupin na zemljevidu je pokazal, da je razdelitev smiselna. Končni AUC na testnih podatkih je znašal 0,79.

• Language: Slovenian
• Keywords: Iskanje skupin v podatkih, gručenje, strojno učenje
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Publisher: [B. Dolenc]
• Year: 2015
• Number of pages: 42 str.
• PID: 20.500.12556/RUL-72418
• COBISS.SI-ID: 1536527811
• Publication date in RUL: 16.09.2015

Licences

• License: CC BY 2.5 SI, Creative Commons Attribution 2.5 Slovenia
• Link: https://creativecommons.org/licenses/by/2.5/si/deed.en
• Description: You are free to reproduce and redistribute the material in any medium or format. You are free to remix, transform, and build upon the material for any purpose, even commercially. You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

Secondary language

• Language: English
• Title: Geographic segmentation of users and its use in advertising

Abstract

In modern web advertising the goal is not only deliver an ad to a broad number of customers, but to target particular customers who are more likely to be interested in content. If the user location is known, we can estimate click on ad based on previous visitors. The company Zemanta recognized the need for geographic audience segmentation, and they have invited students to solve their challenge. The goal was geographic segmentation of web pages visitors based on the ZIP code they come from and development of a prediction model, which can estimate the probability of click on the ad, with accuracy (AUC score) around 0,75. In this dissertation, we describe our the solution to the challenge. Our user segmentation identified 20 groups. There were large differences between them considering population density, urbanization and other demographic indicators. Plotting results on map revealed, that segmentation is meaningful. Our final AUC score on test data was 0,79.

• Keywords: Clustering, Data mining, Machine learning