Using Neural Networks to reduce noise in internet of things data streams
dc.contributor.author | Magondu, Samuel | |
dc.date.accessioned | 2017-01-05T06:53:42Z | |
dc.date.available | 2017-01-05T06:53:42Z | |
dc.date.issued | 2016-11 | |
dc.identifier.uri | http://hdl.handle.net/11295/99015 | |
dc.description.abstract | Noise in the Internet of Things is threatening to drown out sensor data. The problem is growing as more and more devices are being connected to the internet. The noise comes from the electric components both within and without the IoT devices. Other sources of noise include poor calibration. There is thus a need to ensure accurate data is collected in a cost effective way as noisy data might prove disastrous. This study sought to find out the suitability of using neural networks as a filter and also compared its performance to a Kalman filter. An Artificial Neural Network filter application was developed using rapid application prototyping using simulated data to test. The results showed that the Artificial Neural Network filter was reliable to filter out the noise compared to other filtering solutions such as the Kalman filter. Despite the Artificial Neural Network being about 15 times slower than the Kalman filter, it was found to be more accurate. It was thus found that an Artificial Neural Network is much more accurate than a Kalman filter and makes a good noise filter for IoT devices. | en_US |
dc.language.iso | en | en_US |
dc.publisher | University of Nairobi | en_US |
dc.title | Using Neural Networks to reduce noise in internet of things data streams | en_US |
dc.type | Thesis | en_US |