A deep recurrent neural network approach to learn sequence similarities for user-identification

2022 | Decision Support Systems | Citations: 0

Authors: Vamosi, Stefan; Reutterer, Thomas; Platzer, Michael

Abstract: The evolving digital economy entails multifaceted behavioral tracking data such ... Expand

Abstract: The evolving digital economy entails multifaceted behavioral tracking data such as internet clickstreams, loca­ tion trajectories, or taste preferences revealed by music or video streaming. Organizations are increasingly interested in using such data streams to profile customers based on their behavioral similarities for targeting purposes. However, measuring similarities in sequential data is a challenging task. We present a generic deep neural-network-based framework for quantifying the similarity of ordered sequences in observed event histories. This novel approach combines a specific type of recurrent neural nets with a triplet loss cost function used for network training. It yields an embedding space that serves as a similarity metric for complex sequential data, can handle multivariate sequential data and incorporate covariates. We empirically validate the derived similarity metric for user embeddings in the domain of re-identifying users in web browsing histories. We demonstrate its superior performance in discriminating users based on their behavioral browsing patterns by benchmarking against more conventional approaches to measure sequence similarity. In addition, we show that the method­ ology can be used for clustering sub-sequences and re-classifying users based on their observed clickstream behavior. Finally, we critically reflect benefits and possible downsides of the proposed framework, discuss ex­ tensions and promising future applications. An open-source reference implementation can be obtained from github.com/vamosi/tl_rnn. Collapse

Topics: user behavior internet technology open source computer hardware fraud detection

Methods: cluster analysis artificial neural network computational algorithm experiment term frequency–inverse document frequency

Theorizing Routines with Computational Sequence Analysis: A Critical Realism Framework

2022 | Journal of the Association for Information Systems | Citations: 6

Authors: Zhang, Zhewei; Lee, Habin; Yoo, Youngjin; Choi, Youngseok

Abstract: We develop a methodological framework to develop process theories on routines b ... Expand

Abstract: We develop a methodological framework to develop process theories on routines by leveraging large volumes of digital trace data following critical realism principles. Our framework begins with collecting and preprocessing digital trace data, corresponding to the empirically observed experience of critical realism. In the second and third steps of the framework, we identify a finite set of similar repetitive patterns (routines) through computational analysis. We accomplish this by combining frequent subsequence mining and clustering analysis to transform empirical observation into a set of routines that correspond to actual happening in critical realism. Then, we employ a retroduction approach to identify generative mechanisms of the routines. In the final step, we validate the generative mechanisms by evaluating proposed processual explanations and/or eliminating alternatives. We provide an illustrative example of developing a process theory in relation to the collaboration pattern in Wikipedia.Suprateek Sarker was the accepting senior editor. This research article was submitted on April 17, 2017 and underwent three revisions.Zhewei Zhang is an assistant professor in the Information Systems Management group at Warwick Business School, University of Warwick. He received his PhD at Temple University (2016). His research primarily focuses on digital innovation, particularly the organizational and societal impacts brought by the adoption of emerging technologies and innovation through collaboration in nontraditional organizational settings. He is experienced in applying computational approaches, such as sequence analysis and machine learning, to support his research. He served as the co-investigator of several funded projects, such as Innovate UK and GCRF, leading the development of artificial intelligence solutions.Habin Lee is a chair in digital business analytics at Brunel Business School, Brunel University London. He obtained his PhD and MEng in management engineering and management science respectively at KAIST (Korea Advanced Institute of Science and Technology). His research interests include computational social science, digital governance, data analytics for policy modeling and simulation, and sustainable operations management. He published articles in Collapse

Topics: Wikipedia online community knowledge representation computer supported cooperative work information technology infrastructure

Methods: digital trace data cluster analysis hierarchical clustering computational algorithm theory development

Session stitching using sequence fingerprinting for web page visits

2021 | Decision Support Systems | Citations: 1

Authors: De Smedt, Johannes; Lacka, Ewelina; Nita, Spyro; Kohls, Hans-Helmut; Paton, Ross

Abstract: The nature of people’s web navigation has significantly changed in recent years. ... Expand

Abstract: The nature of people’s web navigation has significantly changed in recent years. The advent of smartphones and other handheld devices has given rise to web users consulting websites with more than one device, or using a shared device. As a result, large volumes of seemingly disjoint data are available, which when analysed together can support decision-making. The task of identifying web sessions by linking such data back to a specific person, however, is hard. The idea of session stitching aims to overcome this by using machine learning inference to identify similar or identical users. Many such efforts use various demographic data or device-based features to train matching algorithms. However, often these variables are not available for every dataset or are recorded differently, making a streamlined setup difficult. Besides, they often result in vast feature spaces which are hard to use for actionable interpretation. Collapse

Topics: web page electronic commerce privacy marketing management decision making

Methods: machine learning computational algorithm random forest classification sequence mining experiment

Changes in emergent software development routines: The moderation effects of routine diversity

2021 | International Journal of Information Management | Citations: 0

Authors: Deng, Tianjie; Robinson, William N.

Abstract: Research into the role of routine change has been conducted in relatively stable ... Expand

Abstract: Research into the role of routine change has been conducted in relatively stable and structured organizational settings. This study extends the current understanding of routine change to more fluid organizational forms. Drawing on the literature in organizational routines and routine change, we develop and test a model that captures the dynamic relationships among routine change, routine diversity, and project performance in the context of open-source software development. By sequence-mining the digital trace data of OSS projects, we show that routine change reduces project popularity; however, this effect is mitigated as routine diversity in­ creases. Thus, routine diversity provides a coping mechanism for participants to adapt to changing routines. This study reveals the dynamics in open-source development routines and their effects on project popularity, which can apply to various other fluid organizational forms. Collapse

Topics: open source organizational productivity digital dashboard project productivity organizational context

Methods: sequence mining descriptive statistic machine learning longitudinal research ordinary least square

Theories: dynamic capabilities theory

Detecting Structure In Chaos: A Customer Process Analysis Method

2018 | European Conference On Information Systems | Citations: 0

Authors: Palvölgyi, Elisabeth Z

Abstract: Detecting typical patterns in customer processes is the precondition for gaining ... Expand

Abstract: Detecting typical patterns in customer processes is the precondition for gaining an understanding about customer issues and needs in the course of performing their processes. Such insights can be translated into customer-centric service offerings that provide added value by enabling customers to reach their process objectives more effectively and rapidly, and with less effort. However, customer processes performed in less restrictive environments are extremely heterogeneous, which makes them difficult to analyse. Current approaches deal with this issue by considering customer processes in large scope and low detail, or vice versa. However, both views are required to understand customer processes comprehensively. Therefore, we present a novel customer process analysis method capable of detecting the hidden activity-cluster structure of customer processes. Consequently, both the detailed level of process activities and the aggregated cluster level are available for customer process analysis, which increases the chances of detecting patterns in these heterogeneous processes. We apply the method to two datasets and evaluate the results’ validity and utility. Moreover, we demonstrate that the method outperforms alternative solution technologies. Finally, we provide new insights into customer process theory. Collapse

Topics: process mining email spam information storage and retrieval system business process management

Methods: computational algorithm sequence mining cluster analysis design science descriptive statistic

Theories: dominant logic

Consumer-Oriented Tech Mining: Integrating the Consumer Perspective into Organizational Technology Intelligence - The Case of Autonomous Driving

2017 | HICSS | Citations: 0

Authors: Egger, Marc; Schoder, Detlef

Abstract: To avoid missing technological opportunities and to counteract risks, organizati ... Expand

Abstract: To avoid missing technological opportunities and to counteract risks, organizations have to scan and monitor developments in the external environment through a structured process of technology intelligence. Previous approaches in tech mining—the application of text mining for technology intelligence —have primarily focused on the elicitation of technical or legal information from web, patent, or research databases. However, knowledge of consumers’ needs, fears, and hopes is a prerequisite for the success of an emerging technology in the marketplace. Thus, we claim that technology intelligence needs to also consider consumers’ technology perceptions. Hence, we propose a novel and comprehensive approach to collect user-generated content from the web and apply text mining to derive consumer perceptions. In doing so, we align with an established tech-mining process. This paper illustrates our approach on the emerging technology of autonomous driving and provides an initial indication of concurrent validity. Collapse

Topics: autonomous driving system innovation management consumer behavior web search engine Google+

Methods: survey sentiment analysis support vector machine natural language processing supervised learning

Searching for “Stability” in Fluidity: A Routine-based View of Open Source Software Development Process

2016 | International Conference on Information Systems | Citations: 0

Authors: Li, Xinyu; Yoo, Youngjin; Zhang, Zhewei

Abstract: Open source software communities are fluid form of organizing where the constitu ... Expand

Abstract: Open source software communities are fluid form of organizing where the constituent members and the interactions among the members are continuously and rapidly changing over time. However, certain stability such as identity and norms maintains in spite of the dynamic nature of the communities. This paper seeks to resolve the tension between the fluidity and stability that co-present in open source community through a lens of organizational routine, and proposes a two-step approach to capture routines in the context of open source software development process. Using 200 open source projects on GitHub as a preliminary analysis, this research-in-progress demonstrates the capability of the method while expecting to unleash its whole potential in a future study. Collapse

Topics: JavaScript programming language open source online community open source community

Methods: cluster analysis digital trace data sequence mining descriptive statistic hierarchical clustering

Mining consensus preference graphs from users' ranking data

2013 | Decision Support Systems | Citations: 3

Authors: Chen, Yen-Liang; Cheng, Li-Chen; Huang, Po-Hsiang

Abstract: The group ranking problem consists of constructing coherent aggregated results f ... Expand

Abstract: The group ranking problem consists of constructing coherent aggregated results from preference data provided by decision makers. Traditionally, the output of a group ranking problem can be classified into ranking lists and maximum consensus sequences. In this study, we propose a consensus preference graph approach to represent the coherent aggregated results of users' preferences. The advantages of our approach are that (1) the graph is built based on users' consensuses, (2) the graph can be understood intuitively, and (3) the relationships between items can be easily seen. An algorithm is developed to construct the consensus preference graph from users' total ranking data. Finally, extensive experiments are carried out using synthetic and real data sets. The experimental results indicate that the proposed method is computationally efficient, and can effectively identify consensus graphs. Collapse

Topics: database system

Methods: computational algorithm experiment case study literature study sequence mining

Two-stage database intrusion detection by combining multiple evidence and belief update

2013 | Information Systems Frontiers | Citations: 0

Authors: Panigrahi, Suvasini; Sural, Shamik; Majumdar, Arun K.

Abstract: Insider threats have gained prominence and pose the most challenging threats to ... Expand

Abstract: Insider threats have gained prominence and pose the most challenging threats to a database system. In this paper, we have proposed a new approach for detecting intrusive attacks in databases by fusion of information sources and use of belief update. In database intrusion detection, only intra-transactional features are not sufficient for detecting attackers within the organization as they are potentially familiar with the day-to-day work. Thus, the proposed system uses inter-transactional as well as intra-transactional features for intrusion detection. Moreover, we have also considered three different sensitivity levels of table attributes for keeping track of the malicious modification of the highly sensitive attributes more carefully. We have analyzed the performance of the proposed database intrusion detection system using stochastic models. Our system performs significantly better compared to two intrusion detection systems recently proposed in the literature. Collapse

Topics: database system intrusion detection system information system use online learning system structured query language

Methods: experiment simulation anomaly detection machine learning Markov chain Monte Carlo

Theories: Dempster-Shafer theory theory of information warfare

A prediction framework based on contextual data to support Mobile Personalized Marketing

2013 | Decision Support Systems | Citations: 1

Authors: Tang, Heng; Liao, Stephen Shaoyi; Sun, Sherry Xiaoyun

Abstract: Personalized marketing via mobile devices, also known as Mobile Personalized Mar ... Expand

Abstract: Personalized marketing via mobile devices, also known as Mobile Personalized Marketing (MPM), has become an increasingly important marketing tool because the ubiquity, interactivity and localization of mobile devices offers great potential for understanding customers' preferences and quickly advertising customized products or services. A tremendous challenge in MPM is to factor a mobile user's context into the prediction of the user's preferences. This paper proposes a novel framework with a three-stage procedure to discover the correlation between contexts of mobile users and their activities for better predicting customers' preferences. Our framework helps not only to discover sequential rules from contextual data, but also to overcome a common barrier in mining contextual data, i.e. elimination of redundant rules that occur when multiple dimensions of contextual information are used in the prediction. The effectiveness of our framework is evaluated through experiments conducted on a mobile user's context dataset. The results show that our framework can effectively extract patterns from a mobile customer's context information for improving the prediction of his/her activities. Collapse

Topics: database system mobile system web page mobile phone

Methods: computational algorithm experiment association rule learning simulation machine learning