This is for someone looking to solve Timeseries forecasting in java.It is a miniature version of the ARIMA forecast in R ( miniature becasue it does not give you enough flexibilites unlike R) but solves the purpose of forecasting without assuming any hyperparameters for ARIMA.

This simply means just give me timeseries and the period of repition, it will give you the forecast.

If you don't understand period of forecast , look at this : https://robjhyndman.com/hyndsight/seasonal-periods/

What I have done new in this repo is:

I have implemented ARIMA AUTO FIT method like this here https://www.rdocumentation.org/packages/forecast/versions/8.5/topics/auto.arima in java

Timeseries-Forecast using ARIMA + auto fit arima

This repository is forked from https://github.com/Workday/timeseries-forecast where Workday has writen ARIMA forecst in java. But he did not write the rich Auto fit algorithm and instead demanded (p, d, q), (P,D,Q,m) from the user. But here you give only timeseries and m and you are done.

To understand (p, d, q), (P,D,Q,m) : got through https://otexts.com/fpp2/arima.html ( very nice book)

All you need to care about, a time series , m , seasonal:

Time-Series : make sure the timeseries is anomaly free- ARIMA can't deal with anomalies. seasonal- if the data is seasonal, put this falg as true , otherwise false example- data is stock price for every hour ( seasonality can be assumed to be true and m = 24) m - period of seasonality

Timeseries-Forecast

This is a Java open source library which provides a time series forecasting functionality. It is an implementation of the Hannan-Rissanen algorithm for additive ARIMA models. This library is published by the Workday's Syman team, and is used to support basic timeseries forecasting functionalities in some of the Workday products.

How to Use

In order to use this library, you need to provide input timeseries data as well as ARIMA parameters. The ARIMA parameters consist of a) non-seasonal parameters p,d,q and b) seasonal parameters P,D,Q,m. If D or m is less than 1, then the model is understood to be non-seasonal and the seasonal parameters P,D,Q,m will have no effect.

import com.workday.insights.timeseries.arima.Arima;
import com.workday.insights.timeseries.arima.struct.ForecastResult;

// Prepare input timeseries data.
double[] dataArray = new double[] {2, 1, 2, 5, 2, 1, 2, 5, 2, 1, 2, 5, 2, 1, 2, 5};

// Set ARIMA model parameters.
int max_p = 5;
int max_d = 2;
int max_q = 5;
int max_P = 2;
int max_D = 1;
int max_Q = 2;

//default seasonal is true , default maxIterations is 100.

final ArimaParams paramsForecast = ArimaAutoFit.autoFit(anomFreeMetrics, maxp, maxd, maxq, maxP, maxD, maxQ, m, seasonal, maxIterations);

// estimate ARIMA model parameters for forecasting
final ArimaModel fittedModel = ArimaSolver.estimateARIMA(
        paramsForecast, data, data.length, data.length + 1);

final ForecastResult forecastResult = fittedModel.forecast(forecastSize);


// You can obtain upper- and lower-bounds of confidence intervals on forecast values.
// By default, it computes at 95%-confidence level. This value can be adjusted in ForecastUtil.java
double[] uppers = forecastResult.getForecastUpperConf();
double[] lowers = forecastResult.getForecastLowerConf();

// You can also obtain the root mean-square error as validation metric.
double rmse = forecastResult.getRMSE();

// It also provides the maximum normalized variance of the forecast values and their confidence interval.
double maxNormalizedVariance = forecastResult.getMaxNormalizedVariance();

// Finally you can read log messages.
String log = forecastResult.getLog();

How to Build

This library uses Maven as its build tool.

// Compile the source code of the project.
mvn compile

// To generate javadocs
mvn javadoc:javadoc

// To generate a site for the current project
mvn site

// Take the compiled code and package it
mvn package

// Install the package into the local repository, which can be used as a dependency in other projects locally.
mvn install

Dependencies

The library has the following dependencies:

JUnit 4.12, jama

Authors

Here is the Contributors List for the timeseries-forecast library + me ( Amandeep Arora) Please note that the project was developed and ported from an internal repository. Therefore, the commit record does not reflect the full history of the project.

License

Copyright 2017 Workday, Inc.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.