A VSM-Based and Quality-Aware Query Processing Protocol for Wireless Sensor Networks
Qingchun Ren, Qilian Liang · 2007
In this paper, extending our previous works on query processing, we propose two methods to substitute cosine measure for vector similarity: cosine-length measure (CLM) and joint-deference measure (JDM). Through considering the impact of vector length on vector similarity, CLM alleviates the disadvantage of traditional vector space model (VSM), in which the confidence of query answer is degraded since truly similar nodes cannot be elected according to users' requirement. JDM upgrades the accuracy and degrades the complexity for the computation on similarity coefficient through simplifying the measure from vector domain to scalar domain. In addition, with the distributions of measurement error and environment noise known and/or unknown respectively, we formulate the criteria to determine the optimum sample size to meet users' pre-specified target precision. Through simulation, we checked the validity and sensitivity of cosine measure, CLM and JMD methods on answer quality and network lifetime. Furthermore, our simulation results, in this paper, form a set of criteria for method selection basing on specific applications.