| Title |
Latentinio Dirichlė paskirstymo modelio taikymas interneto reklamos vartotojų segmentavimui / |
| Translation of Title |
User segmentation for online advertising using latent Dirichlet allocation. |
| Authors |
Aliulis, Darius |
| Full Text |
|
| Pages |
187 |
| Keywords [eng] |
latent Dirichlet allocation ; online advertising ; user segmentation |
| Abstract [eng] |
Online advertising is a fast growing multi-billion industry. Its revenue has totaled 30.7 billion Euro in Europe and 49.5 billion US dollars in USA in 2014. When to compared to other forms of adverting, online is the most effective means to reach the right customer, but the problem of serving the right ad to the right customer remains. Behavioral targeting addresses this problem and user segmentation is an essential part of it. Behavioral targeting for online advertising has gained more attention from academic researchers in 2009. The early works employed classical clustering methods for user segmentation. Later it has been shown that topic models are able to capture the semantics of user behavior and outperform the classical methods. However, the experiments were done using relatively small sample sizes and small numbers of segments, therefore it is of interest to conduct thorough research into the effects of user segmentation for behavioral targeting. In this paper Latent Dirichlet Allocation model is used to segment users for online advertising. Tools needed for data preparation, modeling experiments and evaluation of segmentation effectiveness are implemented using Big Data technologies. User segmentation and evaluation of results are carried out on a real-world dataset containing user search query and click-through logs. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
Lithuanian |
| Publication date |
2015 |
