PhD Proposal: A Visual Analytics Approach to Temporal Event Sequence

Recommender systems are being widely used to assist people in making decisions, for example, recommending films to watch or books to buy. Despite its ubiquity, the problem of presenting the recommendations of temporal event sequences has not been studied, which consist of recommended actions and their timings that will lead to the desired outcome based on the history of similar archived records. The main novelty of the approach proposed in this dissertation is that it uses event sequences as features to identify similar records and provide appropriate recommendations. While traditional product recommendations can be described with simple explanations such as customers with attributes like yours also looked at this product or watched this movie, my approach can be summarized by the following statement: based on what happened to customers who started with an event sequence similar to yours, what the sequences of actions and their timings are that might lead to your desired outcome.

My research proposes the first attempt at a visual analytics approach to present and explain recommendations of temporal event sequences. I began by presenting a workflow for temporal event sequence recommendation and I have implemented early prototypes, EventAction and PeerFinder to support the workflow. Results from empirical studies showed that my prototypes can assist users in making action plans and raise users' confidence in the plans. Next, I will develop an asymmetric collaborative framework to support the shared use of my prototypes. Then, I will conduct case studies in multiple domains to demonstrate the effectiveness and safety of generating temporal event sequence recommendations based on personal histories.

Finally, I will summarize design guidelines for the construction of user interfaces for temporal event sequence recommendation and discuss ethical issues in dealing with personal histories.

My dissertation will contribute a systematic workflow, an interactive prescriptive analytics system, and an asymmetric collaborative framework for temporal event sequence recommendation. The empirical studies and case studies will generate interface design guidelines and identify ethical issues, opening new avenues of research in temporal event sequence recommendations based on personal histories.

Examining Committee:

Chair: Dr. Ben Shneiderman

Dept rep: Dr. Neil Spring

Members: Dr. Catherine Plaisant

Dr. Niklas Elmqvist