UPM Institutional Repository

Naive Bayesian decision model for interoperability of heterogeneous systems in an intelligent building environment


Soozaei, Ahmad Shahi (2015) Naive Bayesian decision model for interoperability of heterogeneous systems in an intelligent building environment. PhD thesis, Universiti Putra Malaysia.


The growing number of devices and heterogeneity of systems in intelligent building leads to establish an intelligent interoperability framework among heterogeneous systems in a federated manner. Interoperation complexities are often faced among heterogeneous systems that are data-intensive in nature. Moreover, automated decision making and communication response time are not efficient due to the heterogeneous systems and high load of receiving events lead to systems faults in terms of conflict occurrences. In typical heterogeneous systems, conflicts could be occurred when more than two events are simultaneously activated. In addition, another challenge in interoperability with growth of devices in intelligent building is a poor performance in intelligent building environment that can be a bottleneck and will limit the performance of an intelligent space as well. In such a case, the model encounter with latency prediction, while the goal of the model is to roughly predict and trigger the activated events in right time. As response to the aforementioned problems, many studies have been carried out in the area of Activities of Daily Lives (ADLs). These studies deal with user preferences and intentions as well as activity recognition which is based on processing data obtained through sensors reading that is needed more investigation in lower layers (e.g. sensor layer) of intelligent building. Nevertheless, most researchers of ADLs did not consider the decision support ability in lower layers of intelligent environment in terms of delay in response time and interoperability. Although there are some recent achievements in lower layers which focused on rule-based system namely Event-Condition-Action (ECA) model by providing mutual interoperation and decision support among heterogeneous systems, it still does not fulfil the requirements in terms of delay in communication response time, automated decision-making without any external intervention, conflict occurrences and minimizing latency prediction. As part of our findings to improve the state of the art in ECA-based model, an effective method based on Naive Bayesian classifier (NB) has been proposed. In addition, to ensure timely automated decision mechanism, achieving sustainability and efficient interoperability among heterogeneous systems, NB model is integrated with weighted priority scheduling and minimizing latency prediction techniques with the availability of dataset from five systems. The proposed approach consists of offline and online phases. In the offline stage, after the preprocessing step, Naive Bayesian model is created. In the online stage, the NB model along with conflict resolution and minimizing latency prediction methods is performed to formulate an efficient interoperability decision model and trigger the related system based on receiving events through the server using the XML SOAP protocol and web services. A comprehensive experimental study is carried out to investigate the effectiveness of interoperability decision model among five available heterogeneous systems. For this purpose, testing of the approach was done in Local Area Network (LAN) setting which interwoven with XML SOAP protocol and web services. Experimental results show a superior effectiveness of proposed approach in comparison with the previous study.

Download File

FSKTM 2015 13RR.pdf

Download (1MB) | Preview

Additional Metadata

Item Type: Thesis (PhD)
Subject: Heterogeneous computing
Subject: Intelligent buildings
Call Number: FSKTM 2015 13
Chairman Supervisor: Md.Nasir Bin Sulaiman, PhD
Divisions: Faculty of Computer Science and Information Technology
Depositing User: Haridan Mohd Jais
Date Deposited: 25 Aug 2017 03:52
Last Modified: 25 Aug 2017 03:52
URI: http://psasir.upm.edu.my/id/eprint/57120
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item