- 🐛 Plugin resamples on columns when they should be interpreted as categorical columns
- Updated code-env descriptor for DSS 12
- 🪲 Fix the bug that was adding an extra date at the end after resampling when the last input timestamp was exactly at the end of a period (week, month, half-year, year)
- 🐛 Keep the empty values rather than filtering them with the extrapolation method "Don't extrapolate (impute nulls)"
- ✂️ Add the extrapolation method "Don't extrapolate (no imputation)" to filter missing values
- 📝 Edit plugin.json to reflect the changes made in 2.0.0
- 📈 Decompose the time series into trend, seasonal and residuals using Seasonal and Trend decomposition using Loess (STL)
- Improve long format options
- ✌️ Handle multiple identifiers
- 🔢 Allow for numerical columns as time series identifiers
- The plugin no longer supports Python 2.7. Yet, its previous recipes, namely Resampling, Windowing, Interval extraction, and Extrema extraction are still running with a Python 2.7 code env.
- 📅 Add more frequencies: business days, quarters, semi-annual
- 💭 Impute categorical values during interpolation and extrapolation
- 🪲 Bugfix: prevent the recipe from occasionnaly adding an empty row at the end of the output dataset
- 🪲 Bugfix: when we set acceptable deviation = 0, minimal segment duration = 0, the first row belongs to the interval and the second does not, the first row is no longer missing from the interval
- 🪲 Fix the bug occurring with weekly, monthly and annual time series which failed to convert to offsets
Add visual recipes to prepare time series data
- 📉 Resample time series data
- 📅 Choose frequencies from nanoseconds to years
- 🔖 Compute aggregations or filter a time series using a rolling window.
↔️ The window size can vary from nanoseconds to years
- 🗻 Extract values around an extremum
- ✂️ Identify intervals or segments of the time series where the values fall within a given range
- ☝️ Handle long format with a unique identifier
- 🐍 Support Python 2.7 and Python 3.6