ARMA model
In my journey of Data Science, I have been studying various Forecasting Models. While using the Autoregression model and Moving Average model, I found out that they are not much reliable and I was not finding a satisfactory result. So I continued to search for new models, But they were seeming to be very complex ones. And then I found out that I can use models that I have used previously, but this time instead of using them separately, I can use the power of Autoregression and Moving Average together. This is known as ARMA (AutroRegression Moving Average )model. Nothing complex about the name.
What is ARMA?
ARMA is the combination of the autoregression and moving average. This means that the model can predict its future values on the basis of past values and also calculate the errors in the past values which will affect future values. This makes the ARMA work far better than the autoregression or moving average solely.
Usage of ARMA
ARMA model is mostly used for the systems that have the function of series. And these series are easily affected by the external unknown factors. The best-known example is the stock market. The stock market is clearly a function of series and is easily affected by the current news.
How to use ARMA?
ARMA has two parameters related to it. Let's call them p and q. p represents the interval value of autoregression. q represents the interval value of the moving average. We need to use the smallest value of p and q with the most efficient. And to measure the performance of the forecasting model, I came across the new term Akaike Information Criterion (AIC) which is offered by the ARMA model. So we can easily compare different models with different p and q values and choose the p and q value best AIC value.