Find in Library

In today’s world of large volumes of data, where the usual statistical estimation methods are commonly inefficient or, more often, impossible to use, aggregation methodologies have emerged as a solution for statistical inference. This work proposes a novel procedure for time series regression modelling, in which maximum entropy and information theory play central roles in the replication of time series, estimation of parameters, and aggregation of estimates. The preliminary results reveal that this three-stage maximum entropy approach is a promising procedure for time series regression modelling in big data contexts.