University of Twente Student Theses

Login

Inference of social networks in office buildings based on sensor data from lighting systems

Lingen, R.F.J. van (2017) Inference of social networks in office buildings based on sensor data from lighting systems.

[img] PDF
1MB
Abstract:The aim of this thesis is to investigate to what extend it is possible to infer a social network based on sensor data from lighting systems in an office building. The first part of this thesis is devoted to the development of a model with which we can generate sensor data. We have two reasons for generating our own data. The first reason is that there is no data available for this project. However, the main reason for generating our own data is the fact that we want to control the input. This enables us to evaluate the performance of the inference methods we will present. We will give a brief overview of literature on the simulation of occupancy in office buildings. Next, we will present a model with which sensor data will be generated. In our model, the main focus will be on how we can incorporate the input (a representation of the interaction amongst the occupants) properly into the process to generate sensor data. We will make use of a graph in which all possible meetings are represented by vertices. Two meetings will be connected if they share an occupant and hence cannot be scheduled together. By generating a sequence of maximal independent sets of this graph, we can generate a meeting schedule. The mechanism to generate this sequence will make use of the interaction amongst the occupants. Using this meeting schedule, we will generate sensor data. The second part of this thesis is devoted to the analysis of the sensor data. We will present three examples of an office building with six occupants that will be used to evaluate the performance of the inference methods. Next, we discuss the main challenge for the inference that arises from the data we have generated. We will propose a series of Bayesian inference algorithms. Each algorithm we present makes better use of the data and its structure. The concept of expectation maximization will be discussed. We will prove that the update rule in the last algorithm we present is exactly the update rule we would get if we would use expectation maximization. Using one year of data, we were able to reconstruct the three social networks with a maximal absolute error of 0.0013. Finally, we will apply the model to generate sensor data and the last inference algorithm to an example of an office building with twelve occupants and we analyze the results. Using one year of data, we were able to reconstruct the social network with a maximal absolute error of 0.0001 for this example. Keywords: interaction, occupants, sensor data, maximal independent set, Bayesian inference, expectation maximization
Item Type:Essay (Master)
Clients:
Philips Lighting, Eindhoven, The Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:31 mathematics
Programme:Applied Mathematics MSc (60348)
Link to this item:https://purl.utwente.nl/essays/74155
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page