A group of researchers from the Massachusetts Institute of Technology (MIT) and the University of California, Berkeley have developed a system that allows nonexperts to generate predictions in just a few seconds.
The time series predict database (tspDB) integrates prediction functionality atop an existing time-series database.
MIT's Abdullah Alomar said a novel time-series-prediction algorithm is key to tspDB's success, and is particularly effective at basing predictions on multivariate time-series data, as well as at estimating its volatility to generate a confidence level for its predictions.
The multivariate singular spectrum analysis (mSSA) algorithm can forecast future values and fill in missing data points with greater accuracy and efficiency than state-of-the-art deep learning techniques.
Alomar said, "One reason I think this works so well is that the model captures a lot of time series dynamics, but at the end of the day, it is still a simple model."
From MIT News
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found