Integrated discovery of location prediction rules in mobile environment

Pattern-based prediction is one of the widely used approaches to predict the future location of the users in a mobile environment. Currently, pattern-based prediction is performed in two sequential steps: discovering a set of sequential frequent patterns, followed by generating the prediction rules. However, existing methods cannot forecast locations where their support is less than the threshold. Therefore, some useful patterns with low support cannot be discovered which leads to the reduction in the prediction power. This problem mainly comes from applying a two-step sequential approach. This paper discusses this problem and proposes a novel integrated framework for generating the pattern-based prediction rules. It divides database such that each location has a separate partition. Then at each partition, it directly discovers the prediction rules for the corresponding location through applying a local support threshold. To our best knowledge, this is the first work which integrates the mining and prediction steps instead of applying the sequential approach. Through experimental evaluation considering different conditions, our proposed technique demonstrates more accurate and efficient results than the sequential forecasting scheme.