Filtern
Dokumenttyp
- Dissertation (2) (entfernen)
Sprache
- Englisch (2) (entfernen)
Schlagworte
- Nutzerverhalten (2) (entfernen)
Institut
- Fakultät Wirtschaftswissenschaften (2) (entfernen)
Analysis of User Behavior
(2020)
Online behaviors analysis consists of extracting patterns from server-logs. The works presented here were carried out within the "mBook" project which aimed to develop indicators of the quantity and quality of the learning process of pupils from their usage of an eponymous electronic textbook for History. In this thesis, the research group investigates several models that adopt different points of view on the data. The studied methods are either well established in the field of pattern mining or transferred from other fields of machine learning and data mining. The authors improve the performance of archetypal analysis in large dimensions and apply it to unveil correlations between visibility time of particular objects in the e-textbook and pupils' motivation. They present next two models based on mixtures of Markov chains. The first extracts users' weekly browsing patterns. The second is designed to process essions at a fine resolution, which is sine qua non to reveal the significance of scrolling behaviors. The authors also propose a new paradigm for online behaviors analysis that interprets sessions as trajectories within the page-graph. In this respect, they establish a general framework for the study of similarity measures between spatio-temporal trajectories, for which the study of sessions is a particular case. Finally, they construct two centroid-based clustering methods using neural networks and thus lay the foundations for unsupervised behaviors analysis using neural networks.
Online advertising has become one of the most important dimension of corporate communications. In recent years, a new form of advertising on the Internet has emerged: real-time advertising. Among others, it allows companies to identify potential customers and target them with respect to their interests. In this way, real-time advertising can increase advertising effectiveness and it could, at the same time, improve user experience. With the emerge of this new form of advertising, statistical models have become even more important because they are now being increasingly used to predict online user behavior. The articles included in this dissertation analyze user-level clickstream data generated during multi-channel advertising campaigns (including TV advertising) and during real-time auctions. The goal of the analyses conducted here is to better understand advertising effects and to support decision-making in this context. Most of the analyses are based on Bayesian models. These models allow for a very flexible structure, which enables researchers to model, for instance, heterogeneity across different types of users or non-linear parameters such as users´ reaction times and exponential decay of advertising effects. In addition, these models allow for the inclusion of prior knowledge of parameter distributions, and, therefore, they are well suited for iterative analyses based on clickstream data. Bayesian models can be evaluated in different ways. Instead of only relying on statistical metrics, the articles included in this dissertation aim to estimate the economic value of these models based on their predictive performance. Although this measure can only approximate their true economic value, this approach can be used to compare and evaluate different models and to illustrate the impact of predictive analyses for companies in the context of big data. This dissertation contributes to both information systems research and marketing research and has many managerial implications. First, a process is developed to determine optimal sample sizes representing the best balance between computational costs and predictive accuracy in e-commerce in particular and big data contexts in general. In practice, this process can be used to reduce infrastructure and computational costs. Second, the articles included here describe models that can be used to measure the impact of television ads on users' online shopping behavior. The models can provide insights concerning the effectiveness of individual television ads, the interactions between different advertising channels and the difference in user behavior of TV-induced customers and their non-TV-induced counterparts. Thereby, the models could support decision-making with respect to future advertising campaigns and targeting. Third, the articles describe several possibilities to extend and improve decision support systems currently used in e-commerce and marketing. These improvements enable practitioners to predict users´ interests for arbitrary products and services by using corresponding websites as dependent variables. This approach can be used to improve the effectiveness of real-time advertising campaigns, especially those intended to raise brand awareness among customers.