Speaker
Description
The adoption of Internet of things (IoT) across different business domains enables smart connection between the physical world and the digital world. Voluminous amounts of data have been produced in the various business context such as manufacturing, logistics, transportation, retail, and healthcare, since the past decade as the miniaturisation of IoT devices increases (Marjani et al. 2017). The ubiquitous sensors and tracking devices installed on smartphones have been utilised as basic IoT infrastructure to support these distinct applications, reshaping human behaviours as well as the relationships and interaction patterns between human, things, and the environment.
The human trajectory data generated from IoT can significantly deepen the understandings on the human behaviour in the business context by revealing the humans’ interaction to the actual world. As IoT captures how human interact with the physical world in certain time and location (Lee and Lee 2015), the trajectory data generated from IoT is offering new way of understanding the human behaviour to create the value. In the age of IoT, humans leave an easily traceable digital footprint not only when they visit a website online, but also when they behave in the physical world. So firms are more interested in capturing the digital footprints of their consumers to understand and predict human behaviour than ever before (Guha and Kumar 2018).
Behavioural sequence prediction using IoT data is of value to both researchers and practitioners in business context. For the researchers, data from direct observation can unveil the hidden feature of customer behaviour which has been mainly visited and viewed through traditional positivism approach. Practitioners in the firm also can take a great opportunity to provide very proactive services by coping with future possible customer behaviour. Despite these merits of behavioural sequence prediction, the relevant research is still lacking due to the difficulty of data collection and lack of analytic approach (Abedi et al. 2014).
In particular, leveraging behavioural trace data in real world to implement the prediction model is challenging for two reason. Firstly, finding optimal size of training and prediction points after certain observation is really difficult. How much learning and observation are required for reasonable prediction performance is not only the technical problem in terms of building prediction model, but also the economic problem as all these have something to do with the cost for prediction. Secondly, extracting valuable information by fully harnessing behavioural trace data is another challenge. The trace data collected from IoT tends to have very simple form of ‘stamp’ or ‘sequence’ data, so it requires further processing to be used for complex prediction task. Therefore, how to discover the meaningful information behind the simple trace data is a critical key to build a successful prediction model.
In this research-in-progress paper, we formalise the behavioural sequence prediction problem to tackle and clarify these two challenges within the exhibition context, which is very natural setting without any external treatment such as marketing purpose but can be easily transferrable to actual business scenario. To resolve the formalised prediction problem, we suggest the novel analytic approach by utilising the information extracted from the IoT generated data - frequent sequences, association network among exhibition booth, and geo-network. With these models, we will try to investigate the economic natures of the prediction model by examining the relationship among ‘economic variables in learning-based approach’ such as size of training data, size of input for prediction (i.e., observation points), and depth of prediction, thereby expecting to verify the existence of trade-off relationship among the variables and the phenomena of diminishing marginal accuracy under the training size and depth of prediction.
References
Lee, I., and Lee, K. 2015. “The Internet of Things (IoT): Applications, Investments, and Challenges for Enterprises,” Business Horizons (58:4), Elsevier, pp. 431–440. (https://doi.org/10.1016/J.BUSHOR.2015.03.008).
Guha, S., and Kumar, S. 2018. “Emergence of Big Data Research in Operations Management, Information Systems, and Healthcare: Past Contributions and Future Roadmap,” Production and Operations Management (27:9), John Wiley & Sons, Ltd (10.1111), pp. 1724–1735. (https://doi.org/10.1111/poms.12833).
| Keywords | IoT, behavioural analysis, people analytics |
|---|
