System and Method of Forecasting Based on Periodical Sketching of Big Data of Observations

A method of forecasting comprises, for each respective time period of a plurality of time periods, transforming a plurality of observations from the respective time period into observation sketches. The method then proceeds to extract a set of targeted attributes and constraints from a forecast request, and construct a set expression using the targeted attributes and the associated constraints. The method further comprises selecting one or more sample periods from the plurality of time periods, and for each sample period of the one or more selected sample periods: mapping the set of targeted attributes to observation sketches in the sample period to identify a subset of sketches, and performing a set operation on the subset of sketches using the set expression to determine a cardinality corresponding to the sample period; and generating a forecast result using one or more cardinalities corresponding to the one or more sample periods.