Hidden markov model for decision making among heterogeneous systems in intelligent building

• Main Author: Abba, Babakura
• Format: Thesis
• Language: English
• Published: 2014
• Online Access: http://psasir.upm.edu.my/id/eprint/64117/1/FSKTM%202014%2029%20edited.pdf; http://psasir.upm.edu.my/id/eprint/64117/
• Institution: Universiti Putra Malaysia

The idea of intelligent building promises the ability to automate the environment by installing the needed devices for controlling context aware, personalized, adaptive and anticipatory services. Intelligent building can in this way be referred to a term normally used to characterize a building that incorporates technology and services through networking to improve power efficiency and enhance the nature of living. The inability of systems, devices and sensors to interoperate is the main drawback in intelligent building. They operate at different platform, different configuration and different languages. Hence it is difficult to perform intelligent building operations due to high heterogeneity. The idea behind this study is to design an effective model to resolve the difficulty of decision making among subsystems in a building environment. Existing work done by Perumal et al. (2013) had tackled the problem of interoperation using the Event Condition Action (ECA) mechanism to perform decision making among subsystems. The ECA mechanism uses the rule based to trigger actions and yet the model resulted in poor response time. In order to improve the response time a machine learning algorithm like Hidden Markov Model (HMM) instead of the rule-based is used. HMM is chosen due to the characteristics it possesses such as probabilistic, statistical, machine learning as well as its robustness and scalability has been seen as an efficient and effective model to tackle the problem of interoperation in the intelligent building. We hypothesized that the response time can be improved without sacrificing the system accuracy through machine learning. From our experimentation results showed that HMM managed to reach 95% accuracy on all the data set generated from the pre-defined rule-based and reduced the response time significantly. The model is compared with other selected machine learning such as Naïve Bayes and Fuzzy Logic to show the correctness of the system. The framework of Perumal et al. (2013) was improved by replacing the ECA with the HMM and implementing the framework in the intelligent building.