Tensor Completion based Prediction in Wireless Edge Caching

Abstract

Missing data is inevitable and ubiquitous in Wireless edge caching. A handful of completion methods are present among which the tensor-based models have been shown to be the most advantageous for missing data imputation. Despite their superior imputation accuracies, they are not investigated in wireless edge caching applications. The primary goal of the paper is to improve the prediction via tensor completion methods in Edge caching scenarios. Specifically, since prediction accuracy cannot be improved by an imputation model when the observed training data is sparse, canonical polyadic decomposition-based completion is used to impute the missing data. The simple mean value interpolated completed data is regarded as the baseline approach. For 20–60% sparse data, simulations show that the improvements of prediction error is increased in MovieLens dataset, as the rank of CPD is increased.