ctsa
Univariate ARIMA (Autoregressive Integrated Moving Average)
Emscripten port of the native C package ctsa for univariate time series analysis and prediction.
API
Interface of ctsa
consists of four functions that all take a 1D vector with observations over time.
const ctsa = const diff = ctsa // lag, differencesconst acf = ctsaconst pacf = ctsaconst pred errors = ctsaconst pred errors = ctsa
ARIMA Method (method)
0 - Exact Maximum Likelihood Method (Default)
1 - Conditional Method - Sum Of Squares
2 - Box-Jenkins Method
Optimization Method (optimizer)
0 - Nelder-Mead
1 - Newton Line Search
2 - Newton Trust Region - Hook Step
3 - Newton Trust Region - Double Dog-Leg
4 - Conjugate Gradient
5 - BFGS
6 - Limited Memory BFGS (Default)
7 - BFGS Using More Thuente Method
ACF Method
0 - Default Method
1 - FFT Based method
PACF Method
0 - Yule-Walker
1 - Burg
2 - Conditional MLE (Box-Jenkins)
Web demo
You can try ARIMA online in the Forecast app: https://statsim.com/forecast/.
It uses the original arima
package under the hood and applies random search method to find the best values of p
, d
and q
.