Performing predictive analysis on usage analytics

摘要

Methods for predicting future data based on time-dependent data with increased accuracy include generating resampled datasets from a base dataset having at least one time dependent characteristic. Generating the resampled datasets includes randomly resampling data points in the base dataset to increase a pool of data for predicting future data while at least partially maintaining one or more time dependent characteristics of the base dataset. One or more embodiments apply a modified bootstrapping algorithm to the base dataset to generate the resampled datasets. Predicting the future data includes applying a time series algorithm to the resampled datasets to generate a predicted future dataset with improved accuracy by utilizing the time dependent characteristic maintained in the resampled datasets.

主权项

1. A method of predictive analytics, comprising:
identifying, by at least one processor, a base dataset comprising a plurality of sampled data points describing a set of events, wherein the plurality of sampled data points in the base dataset comprise a chronological order; randomly resampling, by the at least one processor, the plurality of sampled data points to obtain a plurality of resampled data points, wherein randomly resampling the plurality of sampled data points comprises randomly resampling a first sampled data point and, for each subsequent resampled data point from the plurality of resampled data points, randomly resampling a data point from a subset of the plurality of sampled data points that is chronologically at the same time as, or after, a most recent resampled data point preceding the resampled data point; generating, by the at least one processor, a plurality of resampled datasets comprising the plurality of randomly resampled data points while maintaining the chronological order of the plurality of sampled data points; and predicting, by the at least one processor, a set of data points corresponding to future events by applying a time series algorithm to the plurality of resampled datasets.