Neural Network-based Association Mining for Mobility Prediction

This paper shows a ubiquitous middleware for the adaptive distributed object system to consider the mobility support of application services in distributed environment. To ensure the seamless connectivity of a moving client, a prediction based on association mining of mobility patterns is used to perform the replication of services in the next base station. The proposed neuro-association mining is based on Apriori to generate rules for prediction and these rules become nodes for the structure of multilayer perceptron (MLP) to classify the next location and replicate the resource currently serving the mobile client. We present our simulation environment for the ubiquitous middleware and a classification performance evaluation where the proposed algorithm shows more accurate and pro-vide more comprehensive structure compared to neural network based classifiers.