Beamforming analysis using Random Forest classifier
Valkama, Riikka (2021-05-20)
Valkama, Riikka
R. Valkama
20.05.2021
© 2021 Riikka Valkama. Tämä Kohde on tekijänoikeuden ja/tai lähioikeuksien suojaama. Voit käyttää Kohdetta käyttöösi sovellettavan tekijänoikeutta ja lähioikeuksia koskevan lainsäädännön sallimilla tavoilla. Muunlaista käyttöä varten tarvitset oikeudenhaltijoiden luvan.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:oulu-202105208048
https://urn.fi/URN:NBN:fi:oulu-202105208048
Tiivistelmä
Wireless communication has a long history that has changed shape throughout the centuries, from smoke signals to electromagnetic radiation. Data transmission evolution has made worldwide communication possible and has contributed to globalisation. Today, information can be shared in real-time—for example, to the other side of the world. Wireless communication has evolved to the point that real-time plays a vital role, and data loss should not occur.
For efficient wireless data transmission, a beamforming technique has been developed. This is a signal processing technique used in antennas for directional signal transmission or reception. Beamforming includes numerous variations, making the analysis of beamforming challenging. Due to its complex nature, beamforming is attempted to be understood more simply at a higher level, and for that reason, elements are listed that enable the analysis to check whether beamforming succeeded on the radio.
Machine learning is a new trend in different aspects of technology. Problems are aimed to be solved and predicted more efficiently by using suitable machine learning methods. Machine learning enables more precise analysis and error tracking, which are utilised in combination to minimise errors. Furthermore, machine learning has been integrated into various automation systems. This thesis concentrates on analysing the success of beamforming at a high level and aims to automate testing and provide feedback to radio architects who utilise beamforming. For a high-level analysis, a few criteria define the success of beamforming on the radio.
In this thesis, a machine learning pipeline is presented from prepossessing to the final model, and we demonstrate the promising results we have been able to achieve using the random forest classifier. Such promising results make it possible to continue with the beamforming classification and serve as motivation to improve and gather detailed feedback for the end-user.
For efficient wireless data transmission, a beamforming technique has been developed. This is a signal processing technique used in antennas for directional signal transmission or reception. Beamforming includes numerous variations, making the analysis of beamforming challenging. Due to its complex nature, beamforming is attempted to be understood more simply at a higher level, and for that reason, elements are listed that enable the analysis to check whether beamforming succeeded on the radio.
Machine learning is a new trend in different aspects of technology. Problems are aimed to be solved and predicted more efficiently by using suitable machine learning methods. Machine learning enables more precise analysis and error tracking, which are utilised in combination to minimise errors. Furthermore, machine learning has been integrated into various automation systems. This thesis concentrates on analysing the success of beamforming at a high level and aims to automate testing and provide feedback to radio architects who utilise beamforming. For a high-level analysis, a few criteria define the success of beamforming on the radio.
In this thesis, a machine learning pipeline is presented from prepossessing to the final model, and we demonstrate the promising results we have been able to achieve using the random forest classifier. Such promising results make it possible to continue with the beamforming classification and serve as motivation to improve and gather detailed feedback for the end-user.
Kokoelmat
- Avoin saatavuus [32026]