Distributed Decision Fusion in Wireless Sensor Networks for Contextual Experience Sensing and Recommendation Forming in E-Commerce

The distributed decision fusion algorithm for Wireless Sensor Networks (WSNs) is proposed for contextual experience sensing and personalized recommendation in Ecommerce, wherein the multihop relay and Binary Adder Multiple-access Channel (BAMAC) are considered. Providing reliable services in consumer electronics relies on accurate fusion rules, however, with a large number of device nodes being accessed, reducing service latency by efficiently fusing data has undoubtedly become the preferred option to improve the user’s sense of experience. Specifically, we establish the optimum decision fusion rule based on the log-likelihood ratio (LLR) for independent and non-identically distributed multihop relay WSNs under the Binary Symmetric Channel (BSC). However the Channel State Information (CSI) of the multihop relays should be completely aware and the Poisson binomial distribution needs to complete the traversal of all possible received signals. Thus we accomplish the transformation from Poisson binomial distribution to Binomial distribution by using the average probability. Subsequently, we optimize the complexity of the fusion rule by fixing the local decision rule and obtain suboptimal fusion algorithms for the three cases, the ideal local channel, and the relay channel with small or large crossover probability, respectively. Firstly, the idea from the majority-based fusion statistics is successfully borrowed under the ideal local channel assumption. Secondly, the Chair-Varshney (CV) rule is achieved when the intermediate relay transmission link over the edge network is nearly perfect, i.e., the BSC’s crossover probability is sufficiently small. In particular, the majority-based fusion statistic is also obtained when the transmission probability of the local channel is symmetrically distributed. Thirdly, we assume that the relay BSC has a large crossover probability, and the majority-based decision fusion statistic is correspondingly derived. Furthermore, we develop the decision fusion algorithm under the BAMAC for independent identically distributed WSNs. The fusion performance is determined by both Monte Carlo simulation and theory analysis.