Efficient data retrieval for large-scale smart city applications through applied Bayesian inference

Recent years have witnessed the proliferation of worldwide efforts towards developing technologies for enabling smart cities, to improve the quality of life for citizens. These smart city solutions are typically deployed across large spatial regions over long time scales, generating massive volumes of data. An efficient way of data retrieval is thus required, for post-processing of the data-such as for analytical and visualization purposes. In this paper, we propose a data prefetching and caching algorithm based on Bayesian inference, for the retrieval of data in large-scale smart city applications. A brute-force approach is used to determine the optimal weight correction factor in the proposed prefetching algorithm. We evaluate the optimized Bayesian prefetching algorithm against the Naïve and Random prefetch baselines, using both simulated and actual data usage patterns. Results show that the Bayesian approach can achieve up to 48.4% reductions in actual user-perceived application delays during data retrieval.