ARIMA
: stands for Auto Regression Integrated Moving Average. Wiki Link
Here is a JavaScript implementation for the following models from ARIMA
family:
- AR(p).
- MA(q).
- ARMA(p,q).
- ARIMA(p, d, q)
In two different ways of implementation:
- Pure JS Full Documentation | Code
- TensorFlow JS Full Documentation | Code
It provides ARIMA model APIs. The full equation for the non-seasonal arima model is:
where,
-
$B$ is the backshift operator:$By_t=y_{t-1};B\varepsilon_t=\varepsilon_{t-1};B\phi=\phi;B^dy_t=y_{t-d}$ . -
$y_t'$ is the differenced series:$y_t'=y_t-y_{t-1}$ -
$\mu$ is the mean of the differenced series:$\mu=\frac{1}{T}\sum_{t=0}^Ty_t'$ . -
$\varepsilon_t$ is the residual:$\varepsilon_t=y_t-\hat{y_t}$ . -
$\Phi$ are the auto regression weights:$\Phi(B)=(1-B\phi_1-...-B^p\phi_p)$ . -
$\Theta$ are the moving average weights:$\Theta(B)=(1+B\theta_t+...+B^q\theta_q)$ .
It supports working on Master of Things (MoT) platform for serving IoT solutions.
This project was carried out under the supervision of:
, as a graduation project from 9-month professional training program in Artificial Intelligence powerded by the Egyptian Ministry of Communications and Information Technology (MCIT) in cooperation with EPITA in France.