Machine Learning for the Prevention of Injuries in the Construction Industry
(2020) INTM01 20201Innovation Engineering
- Abstract
- The Swedish construction industry is subject to a high rate of occupational injuries, where overload factors are a significant cause. Through Human Activity Recognition, movement data can be collected and analyzed, enabling the identification of harmful movement patterns with the use of machine learning. This study aims to describe the environmental barriers and stakeholder attitudes towards a smart construction helmet which enables this kind of data collection, while evaluating the performance of the supervised machine learning algorithm Random Forest when applying it to movement data. It asks whether collecting movement data violates the privacy of construction workers, or if there are other significant aspects to consider in the... (More)
- The Swedish construction industry is subject to a high rate of occupational injuries, where overload factors are a significant cause. Through Human Activity Recognition, movement data can be collected and analyzed, enabling the identification of harmful movement patterns with the use of machine learning. This study aims to describe the environmental barriers and stakeholder attitudes towards a smart construction helmet which enables this kind of data collection, while evaluating the performance of the supervised machine learning algorithm Random Forest when applying it to movement data. It asks whether collecting movement data violates the privacy of construction workers, or if there are other significant aspects to consider in the adoption process.
Based on a literature review on the Swedish construction industry, digitalization and privacy, interviews were conducted with stakeholders within five relevant roles to gather their attitudes towards the smart helmet. Furthermore, a group of eleven subjects participated in the collection of movement data which was further analyzed with the Random Forest algorithm. Analysis of the interview responses demonstrated a positive attitude from all stakeholders, where technology resistance was an obstacle, while privacy was a less emphasized issue. The movement data analysis showed significant recognition skills after using reviewed methods to manipulate the data. However, the collected dataset was not satisfactory to alone show these results but was complemented by an external dataset. The results indicate that the construction industry may be ready for a smart helmet if the presented gains outweigh the technology resistance and the added weight of the IoTdevice. Further research is however needed to develop the recognition skills to analyze more detailed movement data (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9024031
- author
- Johannesson, Sarah LU and Ögren, Johanna LU
- supervisor
-
- Emil Åkesson LU
- organization
- course
- INTM01 20201
- year
- 2020
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- machine learning, Random Forest, Human Activity Recognition, construction industry, digitalization, privacy
- language
- English
- id
- 9024031
- date added to LUP
- 2020-07-02 12:45:15
- date last changed
- 2020-07-02 12:45:15
@misc{9024031, abstract = {{The Swedish construction industry is subject to a high rate of occupational injuries, where overload factors are a significant cause. Through Human Activity Recognition, movement data can be collected and analyzed, enabling the identification of harmful movement patterns with the use of machine learning. This study aims to describe the environmental barriers and stakeholder attitudes towards a smart construction helmet which enables this kind of data collection, while evaluating the performance of the supervised machine learning algorithm Random Forest when applying it to movement data. It asks whether collecting movement data violates the privacy of construction workers, or if there are other significant aspects to consider in the adoption process. Based on a literature review on the Swedish construction industry, digitalization and privacy, interviews were conducted with stakeholders within five relevant roles to gather their attitudes towards the smart helmet. Furthermore, a group of eleven subjects participated in the collection of movement data which was further analyzed with the Random Forest algorithm. Analysis of the interview responses demonstrated a positive attitude from all stakeholders, where technology resistance was an obstacle, while privacy was a less emphasized issue. The movement data analysis showed significant recognition skills after using reviewed methods to manipulate the data. However, the collected dataset was not satisfactory to alone show these results but was complemented by an external dataset. The results indicate that the construction industry may be ready for a smart helmet if the presented gains outweigh the technology resistance and the added weight of the IoTdevice. Further research is however needed to develop the recognition skills to analyze more detailed movement data}}, author = {{Johannesson, Sarah and Ögren, Johanna}}, language = {{eng}}, note = {{Student Paper}}, title = {{Machine Learning for the Prevention of Injuries in the Construction Industry}}, year = {{2020}}, }