Predicting Wireless sensor readings with Neural network
dc.contributor.advisor | Srirama, Satish Narayana | |
dc.contributor.author | Adekunle, Lukman | |
dc.date.accessioned | 2017-04-26T06:49:21Z | |
dc.date.available | 2017-04-26T06:49:21Z | |
dc.date.issued | 2014 | |
dc.description.abstract | Wireless sensor networks are becoming a part of our daily lives, as they act as a bridge between the physical world and the virtual world.One of the problems encountered by this type of networks while trying to fulfill their goals is the rate of energy consumption. The approach considered in this paper was that of an artificial neural network with the aim of reducing the rate of power consumption and thereby increasing the performance and durability of the network. Support vector machines backed artificial neural model was the best of all models picked. It was then compared with a linear regression model to see if there would be any good reasons to migrate to the this new approach. At the end, it was observed that the chosen network performed slightly above the level of the existing model. The implications of the observed results were that another form of prediction model can replace the existing one or alternated with one another in the process of operation of a wireless sensor network. | |
dc.description.abstract | Juhtmevabadest sensorvõrkudest on saamas osa meie igapäevalust. Tegemist on sillaga füüsilise ja virtuaalse maailma vahel. Üheks probleemiks seda laadi võrkude puhul on aga energia tarbimise määr. Käesolevas lõputöös uuriti lähenemist, kus kasutatakse tehisneurovõrke eesmärgiga vähendada energiatarvet ja seeläbi parendada sensorvõrgu efektiivsust ning vastupidavust. Tugivektormasinatega toetatud tehisneuromudel valiti välja kui parim vaatluse all olnud mudel. Seda mudelit võrreldi lineaarse regressiooni mudeliga, et näha kas väljavalitud mudeli puhul leidub mõjuvaid põhjuseid just seda eelistada. Lõpuks selgitati välja, et uuritava mudeli efektiivsus oli veidi kõrgem kui võrreldaval mudelil. Töö tulemustest järeldub, et olemasolevaid ennustusmudeleid võib asendada alternatiivsetega või kasutada neid vaheldumisi juhtmevaba sensorvõrgu töö käigus. | |
dc.identifier.uri | http://hdl.handle.net/10062/56026 | |
dc.language.iso | eng | |
dc.title | Predicting Wireless sensor readings with Neural network | |
dc.title.alternative | Predicting Wireless sensor readings with Neural network | |
dc.type | Thesis |