ARIMA model

In the journey of becoming a data scientist, I have been studying various forecasting models. And this week I have reached a model which is being widely used by Data scientists all over the world. The techniques which I have studied earlier were being used only until this model emerged as a savior. I am talking about the ARIMA model. It's been the most reliable model and its versions are used with different scenarios which I will be studying in the later weeks.

What is ARIMA?

ARIMA stands for Autoregressive Integrated Moving Average. Here the autoregressive part helps to regress the upcoming values by applying a linear function on the past values. The Integrated part helps to find the difference between past consecutive values to indifferent orders to remove it from the level of changes. The moving average part keeps in check of the errors of the actual value from the predicted value and navigates the upcoming values under a certain range of predictability.

Difference between ARIMA and ARMA

We have already noticed that the only difference that seemed to be noticed is the 'I' which stands for Integrated. It means that it removes the difference between consecutive values in different orders from the predictive model which is not available in the ARMA model which makes it a game far inferior to ARIMA.

Usage of ARIMA

This statistical model is mostly used for the time series model for forecasting the future. Hence you could probably assume that it is mostly used in the stock market, forecasting the weather, and other statistical predictions.

How to use ARIMA?

As we know that ARIMA contains three parts. Hence while using the ARIMA model we need to tweak 3 parameters for controlling the three parts. Let us consider ‘p’ for Autoregression, ‘q’ for moving average and ‘d’ for the difference. So ‘p’ needs to be used for the lag in the predictable value, ‘q’ is the error from of the forecasting value from the actual value and as we know we need to make the time series stationary hence the value of ‘d’ is the minimum value to make the time series stationary. As discussed in the previous blog of the ARMA model here also we evaluate the model using Akaike information criterion (AIC) value which can be easily determined through the model.

Thank you for reading the article. Now as for the bonus point on the ARIMA Model

Visit the original source: RohitOnWeb.tech