Sensing & Modelling Human Behaviour


Going through a paper by Mehrotra [1], a very comprehensive review of what has already been achieved through the analysis of social media and mobile data. Amazing to realise the wide range of behaviours that can be observed from this data, from emotion and mood to political election and earthquake propagation.

The analysis process: user behaviour => passive sensing data / user feedback questionnaires => data mining => intervention

These data are ‘ … allowing researchers to address research questions in completely new ways’. On the top of the outlook for this research field, they set the challenge of combining this quantitative data with the classic qualitative method of social sciences. Indeed, the quantitative data is large scale and passive but often hard to interpret. Classic qualitative methods are seen as a resource to understand the quantitative data.

[1] Abhinav Mehrotra and Mirco Musolesi, 3.18 – Sensing and Modeling Human Behavior Using Social Media and Mobile Data, In Comprehensive Geographic Information Systems, edited by Bo Huang,, Elsevier, Oxford, 2018, Pages 313-319, ISBN 9780128047934,

Notes on IoT Analytics


The concept of web analytics is very well understood and heavily used, mostly powered by Google Analytics. How does it translate to the Internet of Things?

There is much less work done in IoT analytics, tools to understand data created by the IoT [1]. Web analytics can provide detailed analysis of user behaviour, enabling to quickly spot and address interaction problems towards a targeted goal. This is in line with the Lean start up from Eric Reis.

What ‘user interaction’ means in web analytics? Page views and events and there related features such as time on site, bounce rate, content, etc. An event consist in four attributes: category, action, label and value. Data is push to the cloud through protocols such as the Universal Measurement Protocol (UMP). In contrast with the web, the IoT offer is very diverse, domain specific and ‘everyone’ build their own visualisation tools with libraries like D3.js.

Mateusz and colleagues developed Pheme, a tool they use to pipe IoT data  into Google Analytics. They explore 4 use cases with raw data, processed data, user engagement data and multiple devices per user. They reported limits regarding the push of historical data (should not be older than 4hrs) and not in the future (e.g. prediction data).

Paul and colleagues raise the potential of real-time analysis in the context of a Social IoT (SIoT) [2]. What does it mean to combine all these streams of data at the same time? What could be the potential application. The authors provide example with parking lots and home energy.

Acer and colleagues [3] leverage WiFi to search location and state of physical objects, thus generating uniform data IoT analytics from out of a wide spread wireless technology.

[1] M. Mikusz, S. Clinch, R. Jones, M. Harding, C. Winstanley and N. Davies, “Repurposing Web Analytics to Support the IoT,” in Computer, vol. 48, no. 9, pp. 42-49, Sept. 2015.
doi: 10.1109/MC.2015.260
[2] A. Paul, A. Ahmad, M. M. Rathore and S. Jabbar, “Smartbuddy: defining human behaviors using big data analytics in social internet of things,” in IEEE Wireless Communications, vol. 23, no. 5, pp. 68-74, October 2016. doi: 10.1109/MWC.2016.7721744
[3] U. G. Acer, A. Boran, C. Forlivesi, W. Liekens, F. Pérez-cruz and F. Kawsar, “Sensing WiFi network for personal IoT analytics,” 2015 5th International Conference on the Internet of Things (IOT), Seoul, 2015, pp. 104-111. doi: 10.1109/IOT.2015.7356554