Introducing Timeseries Insights API: real-time forecasting and anomaly detection over trillions of events
Businesses are awash in event-driven data, from customers visiting stores or submitting online queries to making purchases or triggering sensors. Leveraging this data for real-time forecasting and anomaly detection is critical to serving customers across changing needs and environments—but doing this kind of time-series analysis in real-time has often required bespoke solutions, saddling businesses with options that are prohibitively expensive and difficult to operate.
To help your business more simply and effectively leverage this data, today we’re announcing the Timeseries Insights API, a highly efficient and scalable system that makes it easier for businesses and developers to quickly gather insights from streamed time series.
The Timeseries Insights API is fully integrated with Google Cloud Storage and Google Cloud PubSub, letting it handle datasets as large as trillions of events. With an efficient “cold-start” ingestion mechanism, the API allows you to load large amounts of historical data to which you can append new events in real-time, and over which you can run continuous forecasting and anomaly detection queries.
So your business can start generating actionable intelligence faster, we designed the API around several principles:
Simple: The Timeseries Insights API has a simple dataset management and query format. Performing forecasting and anomaly detection is as simple as defining training and testing time intervals over the event space and asking the API to highlight any anomalies over one or more event dimensions.
Real-time intelligence: Through its efficient batch ingestion and streaming update interface, the Timeseries Insights API lets you perform real-time analysis over vast amounts of events. Trillions of events can be continuously inspected even as thousands of new events arrive every second.
Scalable: The Timeseries Insights API offers scale for virtually any need. Its backends are deployed in a variety of internal Google applications, where they routinely do hundreds of thousands of queries a second over huge, continuously growing datasets.
Time-series data prediction is critical in a variety of industries. Retailers can analyze inventory and adjust it to seasonality effects, IoT companies can detect intrusion attempts, payment processors can identify promotions from outages, fintech can analyze trending topics etc. We’re excited to see what the API can do for you.