1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Institute of Microengineering and Nanoelectronics / Institut Kejuruteraan Mikro dan Nanoelektronik (IMEN)

Title:	Human activity prediction in smart homes using finite order Markov model and Gaussian distribution
Authors:	Muhammad Raisul Alam (P44359)
Supervisor:	Mamun Ibne Reaz, Assoc. Prof. Dr.
Keywords:	Home automation Markov processes Household electronics Electronic control Gaussian distribution Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia
Issue Date:	18-Mar-2011
Description:	Smart home is an application of ubiquitous computing where the home environment is monitored by ambient intelligence to provide context aware services. The aim of smart home research is to provide smartness to the dwelling facilities for comfort, healthcare, security, safety and energy conservation. Remote monitoring is a common component of smart homes where telecommunication and internet technologies are being used to provide remote home control and to support the resident from specialized assistance centre. In smart home environment, prediction algorithm plays an important role to automate user interactions with the surroundings. A smart home should act proactively according to the resident's requirements based on his previous activities. There are two fundamental problems of prediction in smart homes. One is the prediction of next activity that is going to happen and another is the prediction of approximate time of that activity. The aim of this research is the development of algorithms for activity and temporal prediction in smart homes. A novel algorithm named SPEED (Sequence Prediction via Enhanced Episode Discovery), is proposed for sequential activity prediction which utilized dual state properties of home appliances to extract episodes. Classified episodes are processed and arranged in a finite order Markov model. A method motivated by Prediction by Partial Matching (PPM) algorithm is applied to select the next event from the frequencies of variable length episodes. A temporal prediction algorithm is presented to predict the interval between the activities. The algorithm is based on the hypothesis that smart home event interval follows Gaussian distribution. To predict the starting time of the following activity, it incrementally utilizes the mean and standard deviation of previous history which are applied according to the central limit theory of statistical probability. Further analysis validates the hypothesis that temporal interval follows a Gaussian distribution which was only an assumption previously. Smart home data from MavLab is used to validate the algorithms. MavLab is the testbed of MavHome at University of Texas in Arlington. The data sample consists activities of six inhabitants at MavHome in April 2003. MavHome dataset has 51 different appliances with time and status information. Result shows that, for a fully converged trie the SPEED algorithm exhibits 88.3% prediction accuracy. The temporal interval prediction algorithm shows 90.9% prediction accuracy when verified with a fully converged database. This research developed innovative algorithms for human activity prediction in smart homes using finite order Markov model and Gaussian distribution. The results show reliable outcomes which have significant contribution in activity prediction and anomaly detection.,'Certification of Master'/Doctoral Thesis' is not available,Master of Science
Pages:	97
Call Number:	TK7881.25.A565 2011 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Institute of Microengineering and Nanoelectronics / Institut Kejuruteraan Mikro dan Nanoelektronik (IMEN)

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.