Hi

Can someone please help me to choose which timeseries model I should use for attached data trend.

Thanks

Hi

Can someone please help me to choose which timeseries model I should use for attached data trend.

Thanks

Hi @plarion,

This time series has a seasonal trend, i.e. for each month the trend of the series is repeating. A sudden drop can be seen in Dec month of every year. So, here a model should be used that can take the seasonality into account.

One such model is **Holt Winter’s Model**, which take trend as well as seasonality into account. So, you can use this model for forecasting. One more advanced approach could be to use **ARIMA** model, which also considers the trend and seasonality in the time series.

You can refer the below course to learn more about these models and how to implement them in Python:

Hi @PulkitS,

Thanks for training link. I followed ARIMA model guideline in that article.

However, when I predict, I’m getting straight line as shown in attached picture. i.e the model predict constant value. I’m expecting wavy line.

Thanks

Hi @plarion,

It seems like the ARIMA model was not able to learn the trend and seasonality from the given dataset, which is quite strange to believe. But no problem, you can still try the holt winter’s model or the SARIMAX model on you dataset. If that also does not work, please share your dataset with me if possible and i will try to find the solution.

Thanks @PulkitS. Appreciate your readiness to help community members. I think I’m getting there. However, I don’t have clear solution to describe you.

I suggest a book called Forecasting: Principles and Practice (https://otexts.org/fpp2/) If working on R, it has good step-by-step guidance, especially around exploration of time series data (with R code included)

Hi @plarion,

Since you are getting a straight line for predictions, try altering the parameters of ARIMA model. For instance change value of q (I suppose you would have used 2, make it 4 maybe).