Time series analysis in Java.
Major API changes have been made with release 0.3. The first and most obvious is that the single java-timeseries project has been split up into two seperate modules -- math and timeseries. The timeseries module contains the time series data types, models, and forecasts. The math module contains much of the lower-level supporting structure and is likely to be of less interest to library users.
The arima package has undergone a lot of refactoring and some structural changes. One critical change worth noting is that the arima classes now have the concept of a drift term, and differentiate between the mean and the drift in certain special cases. If clarification is needed, don't hesitate to create an issue for discussion.
For this release, time series linear regression model and forecast classes have been added. However, these are unlikely to be very useful on their own at this time, at least for forecasting purposes. The best you can get out of them currently is a point forecast. For the next release, the time series linear regression forecast class will implement the Forecast interface, which will make it much more useful.
- Seasonal ARIMA models.
- ARIMA forecasting and simulation.
- Random walk simulation and modeling.
- Time series statistics, moving averages, and aggregation.
- Simple, intuitive, and powerful time modeling.
- Autocorrelation function and ACF plot.
Please note that this library is still in early development stages. Major API changes are expected in the near future.
<dependency>
<groupId>com.github.jrachiele</groupId>
<artifactId>timeseries</artifactId>
<version>0.3</version>
</dependency>
compile 'com.github.jrachiele:timeseries:0.3'
Library | Category | License |
---|---|---|
XChart | Graphing | Apache 2.0 |
Smile | Distributions | Apache 2.0 |