Abstract. Online advertising is increasingly switching to real-time bidding on advertisement inventory, in which the ad slots are sold through real-time auctions upon users visiting websites or using mobile apps. To compete with unknown bidders in such a highly stochastic environment, each bidder is required to estimate the value of each impression and to set a competitive bid price. Previous bidding algorithms have done so without considering the constraint of budget limits, which we address in this paper. We model the bidding process as a Constrained Markov Decision Process based reinforcement learning framework. Our model uses the predicted click-through-rate as the state, bid price as the action, and ad clicks as the reward. We propose a bidding function, which outperforms the state-of-the-art bidding functions in terms of the number of clicks when the budget limit is low. We further simulate different bidding functions competing in the same environment and report the performances of the bidding strategies when required to adapt to a dynamic environment.

Behavioral and targeted profiling of users is an important task in marketing and in the advertising industry. Being able to match a given user profile to an advertising that leads to effective purchases is challenging because of a very tiny proportion of users willing to purchase goods and thus monetize the advertising. With such proportions being less than one percent of the overall user population, efficient feature extraction and modeling techniques are required in order to capture and recognize the potential consumers. This paper proposes a new approach for modeling the observed behavior in a mobile advertising platform, where time related features are correlated with additional system level and campaign related performance statistics. We capture the temporal behavior with Hawkes processes and use the estimated parameters as additional features for predicting if a given user profile will be a revenue generating customer.

Video content, of which YouTube is a major part, constitutes a large share of residential Internet traffic. In this paper, we analyse the user demand patterns for YouTube in two metropolitan access networks with more than 1 million requests over three consecutive weeks in the first network and more than 600,000 requests over four consecutive weeks in the second network.In particular we examine the existence of "local interest communities", i.e. the extent to which users living closer to each other tend to request the same content to a higher degree, and it is found that this applies to (i) the two networks themselves; (ii) regions within these networks (iii) households with regions and (iv) terminals within households. We also find that different types of access devices (PCs and handhelds) tend to form similar interest communities.It is also found that repeats are (i) "self-generating" in the sense that the more times a clip has been played, the higher the probability of playing it again, (ii) "long-lasting" in the sense that repeats can occur even after several days and (iii) "semiregular" in the sense that replays have a noticeable tendency to occur with relatively constant intervals.The implications of these findings are that the benefits from large groups of users in terms of caching gain may be exaggerated, since users are different depending on where they live and what equipment they use, and that high gains can be achieved in relatively small groups or even for individual users thanks to their relatively predictable behaviour.

No abstract

In this work, we study YouTube traffic characteristics in a medium-sized Swedish residential municipal network that has ∼ ∼ ∼ ∼ 2600 mainly FTTH broadband-connected households. YouTube traffic analyses were carried out in the perspective of video clip category and duration, in order to understand their impact on the potential local network caching gains. To the best of our knowledge, this is the first time systematic analysis of YouTube traffic content in the perspective of video clip category and duration in a residential broadband network. Our results show that the requested YouTube video clips from the end users in the studied network were imbalanced in regarding the video categories and durations. The dominating video category was Music, both in terms of the total traffic share as well as the contribution to the overall potential local network caching gain. In addition, most of the requested video clips were between 2-5 min in duration, despite video clips with durations over 15 min were also popular among certain video categories, e.g. film videos.