TIMEX: an Automatic Framework for Time-Series Forecasting-as-a-Service

Date: 2022-08-15

Venue: The Sixth International Workshop on Automation in Machine Learning, workshop of the 28th ACM SIGKDD Conference on Knowledge Discovery & Data Mining (KDD), Washington D.C., 2022

Forecasting-as-a-service is a novel and promising research area aiming at designing Cloud-based platforms and systems able to provide users with time-series forecasting ability in a “as-a-service” manner. This paper introduces TIMEX, an automatic “Time eXploration” open-source Python-based framework for time-series forecastingas-a-service. Difefrently from the tools and libraries present in the literature, which provide excellent results only in specific aspects, TIMEX provides a fully automatic end-to-end forecasting pipeline in a “as-a-service” manner comprising data ingestion, data preprocessing, data description, data prediction and service delivery. In addition, w.r.t. the literature, TIMEX introduces relevant algorithmic advances in the automatic characterization of the relationships among time-series, in the automatic definition of the best training window and in the automatic selection of feature transformation. TIMEX has been applied both to the forecasting of the COVID-19 pandemic spread in Italy and to the prediction of two benchmark time-series; results are successfully contrasted with a state-of-theart solution present in the literature.

GitHub repo: link

DOI: waiting for the KDD to publish the proceedings. :(