Skip to content

TIM API

v4.5 [2020-10-23] (current stable release)

New features

  • added in-sample results for RTInstantML​
  • added backtesting results for RTInstantML for fast sampling rates​
  • added model view and accuracy through different forecasting horizons​
  • reworked extended importances to match (non-)daily cycle
  • improved execution time (sampling rate agnostic performance) in anomaly detection
  • added isTimeSpecific flag in the response JSON indicating whether modelZoo is time specific
  • reworked sampling period computation for irregular data with using median

Non-breaking changes

  • removed constraint of backtesting length in RTInstantML
  • holiday column name if do not exist in csv data return warning
  • several bugfixes for irregular data​

Breaking changes

This release does not contain any breaking changes.

v4.4 [2020-08-31]

New features

Non-breaking changes

Breaking changes

This release does not contain any breaking changes but does not support the old Anomaly model from previous versions.

v4.3 [2020-08-06]

New features

Non-breaking changes

Breaking changes

This release does not contain any breaking changes.

v4.2 [2020-04-15]

New features

Non-breaking changes

  • dictionary Simple Moving Average is now applied to all predictors (not just target)
  • added option to use CSV schema for data in JSON requests
  • reworked treemap strings
  • added an option to return normal behavior signal (forecast) in anomaly detection

Breaking changes

This release does not contain any breaking changes.

v4.1 [2020-01-30]

New features

  • authentication via access tokens
  • customizable interpolation of missing values
  • new dictionaries: Identity and Simple Moving Average
  • added option to turn off offsets
  • support for data with regular gaps

Non-breaking changes

  • data doesn't have to be aligned
  • optional HTTP headers X-Client-Name and X-Client-Version
  • changed warnings for extreme values coming to prediction to IQR standard instead of relative difference
  • turned off Piecewise Linearity dictionary for too short datasets
  • turned off Weekdays and Weekrest dictionaries for datasets with too few Mondays
  • fixed bug with predictors that are updated not every day of week
  • not using low quality predictions in anomaly detection

Breaking changes

This release does not contain any breaking changes.

v4.0 [2019-10-15]

New features

  • support for general sampling rates

Non-breaking changes

  • using JSON schemas for request validation (affects error messages)
  • API definition greatly simplified and converted to OpenAPI 3.0 (previously Swagger 2.0)

Breaking changes

  • new licensing system, License-Key HTTP header is required
  • Api-Key HTTP header is required
  • removed "OneTime" usage (deprecated by RTInstantML)
  • removed meteo support
  • refactored anomaly detection features DO