Presence Detection in Homes using Aggregated Sensor Data
(2024) In Master's Theses in Mathematical Sciences FMAM05 20232Mathematics (Faculty of Engineering)
- Abstract
- This thesis presents the development of a binary classifier, which classifies homes as occupied or not. This was done using two different types of neural networks; standard feedforward networks and LSTM-networks. The input to these networks was sensor data collected from devices created by Minut AB. Data from previously used automatic alarm feature was used as ground truth. A major part of the project consisted of preparing and filtering the ground truth and input data. Once this was done a set of suitable hyperparameters was found by tuning the hyperparameters one by one with the other ones fixed. In general the tuning of the hyperparameters was too noisy to make any certain conclusions about which values were optimal.
The classifiers... (More) - This thesis presents the development of a binary classifier, which classifies homes as occupied or not. This was done using two different types of neural networks; standard feedforward networks and LSTM-networks. The input to these networks was sensor data collected from devices created by Minut AB. Data from previously used automatic alarm feature was used as ground truth. A major part of the project consisted of preparing and filtering the ground truth and input data. Once this was done a set of suitable hyperparameters was found by tuning the hyperparameters one by one with the other ones fixed. In general the tuning of the hyperparameters was too noisy to make any certain conclusions about which values were optimal.
The classifiers succeeded in identifying some patterns indicative of home occupancy, outperforming a baseline model, which randomly guesses occupancy status. Despite this, the classifiers' performance did not yield the high accuracy required for their intended applications in heating system regulation and other home automation tasks. The feedforward network model got the best results, but LSTM-networks could potentially be equally good for this task, since the LSTM-networks were trained on smaller amounts of data and data quality appeared to affect the result more than model choice. Areas to improve include preprocessing, the method for choosing hyperparameters and quality of the ground truth data. (Less) - Popular Abstract
- Learn more about how AI could be used to optimize home systems while respecting the privacy of the residents.
In this study, we explore the possibilities of using Artificial Intelligence (AI) to optimize home systems, such as heating. This could potentially save vast amounts of money and also lower the environmental impact from homes. Integrating occupancy detectors with other systems in homes could also make everyday life more comfortable and efficient. It is, however, important to reach a balance between this convenience and privacy, which motivated the chosen approach.
In order to take privacy concerns into account, the data used to create the occupancy detector was sensor data from Minut AB, a company known for their... (More) - Learn more about how AI could be used to optimize home systems while respecting the privacy of the residents.
In this study, we explore the possibilities of using Artificial Intelligence (AI) to optimize home systems, such as heating. This could potentially save vast amounts of money and also lower the environmental impact from homes. Integrating occupancy detectors with other systems in homes could also make everyday life more comfortable and efficient. It is, however, important to reach a balance between this convenience and privacy, which motivated the chosen approach.
In order to take privacy concerns into account, the data used to create the occupancy detector was sensor data from Minut AB, a company known for their privacy-friendly home monitors. Creating something useful from this data seemed to require more than manual identification of connections between the input features, such as humidity and sound level, and whether the space studied was occupied or not. The chosen approach was two types of state of the art AI methods: standard Deep Neural Networks (DNNs) and Long Short-Term Memory (LSTM) networks, which are able to capture long term dependencies in time (needed to, for instance, tell whether someone is at home hours after everyone has gone to bed) without explicitly looking very far backwards in time. Both methods were clearly better than plain guessing, the DNN approach being the most successful, but due to various reasons, such as low quality of parts of the dataset, more work needs to be done to meet the high demands on features in smart home systems.
\\
Although not yet optimal for home system applications, this project demonstrated that more complex AI models, such as LSTM, do not always outperform simpler ones like DNNs as well as helped in laying a foundation for future work creating smart homes with equally smart privacy. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9154181
- author
- Georgson, Melker LU
- supervisor
- organization
- course
- FMAM05 20232
- year
- 2024
- type
- H2 - Master's Degree (Two Years)
- subject
- publication/series
- Master's Theses in Mathematical Sciences
- report number
- LUTFMA-3526-2024
- ISSN
- 1404-6342
- other publication id
- 2024:E9
- language
- English
- id
- 9154181
- date added to LUP
- 2024-06-28 16:02:45
- date last changed
- 2024-06-28 16:02:45
@misc{9154181,
abstract = {{This thesis presents the development of a binary classifier, which classifies homes as occupied or not. This was done using two different types of neural networks; standard feedforward networks and LSTM-networks. The input to these networks was sensor data collected from devices created by Minut AB. Data from previously used automatic alarm feature was used as ground truth. A major part of the project consisted of preparing and filtering the ground truth and input data. Once this was done a set of suitable hyperparameters was found by tuning the hyperparameters one by one with the other ones fixed. In general the tuning of the hyperparameters was too noisy to make any certain conclusions about which values were optimal.
The classifiers succeeded in identifying some patterns indicative of home occupancy, outperforming a baseline model, which randomly guesses occupancy status. Despite this, the classifiers' performance did not yield the high accuracy required for their intended applications in heating system regulation and other home automation tasks. The feedforward network model got the best results, but LSTM-networks could potentially be equally good for this task, since the LSTM-networks were trained on smaller amounts of data and data quality appeared to affect the result more than model choice. Areas to improve include preprocessing, the method for choosing hyperparameters and quality of the ground truth data.}},
author = {{Georgson, Melker}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master's Theses in Mathematical Sciences}},
title = {{Presence Detection in Homes using Aggregated Sensor Data}},
year = {{2024}},
}