Now, let’s consider more than 10 observations. Which model should you use in the case of larger data sets? In the next video, Chiranjoy will explain this in detail.

In the case of more than 10 data points, for using a technique among the smoothing/ARIMA techniques, you need to keep the following points in mind:

- To capture the level in time series data, the simple exponential smoothing technique is used. An example in which the simple exponential smoothing technique is used is the forecasting of the annual GDP of a developed country. In this case, there may or may not be an obvious trend in the data and the seasonal component is absent. Thus, using a simple exponential smoothing technique makes sense here.

- Holt’s exponential smoothing technique / ARIMA method works best in capturing both the level and the trend in the time series data. An example in which Holt’s exponential smoothing technique/ARIMA method can be used is when there is an obvious trend in the data but no specific seasonality exists. For example, the sales data for iPhones for the last 10–12 years show an increasing trend but no specific seasonality; thus, the Holt’s method / ARIMA method can be used to forecast the time series data.

- Now, to capture the level, the trend and seasonality, the Holt-Winters’ exponential smoothing technique / SARIMA works best. However, this method should be used when the dataset has no exogenous variables. An example in which the techniques mentioned here can be used is in determining the number of monthly visitors to an amusement park. With increased marketing, the number of visitors to the amusement park keeps increasing and thus, the data shows a trend. However, these numbers show a seasonal pattern. For example, during the summer and winter vacation months, visitors tend to flock to the amusement park even more, which implies that the dataset has a seasonality component.

- The ARIMAX/SARIMAX method works best for capturing the level, the trend and the seasonality in time series data when some exogenous variables are present. The ARIMAX method may not capture the seasonality but the SARIMAX method does. An example of using these methods is in determining the monthly sales of an e-commerce website. Here, the sales are affected by numerous external factors and thus, the ARIMAX/SARIMAX method will work well here.

You can further test your understanding of the topic by solving the exercises given in the optional practice session. Here is the link to the segment.

Let us now summarise what we have studied in this module in the next segment.