time series modelling



For discussions related to modeling, machine learning and deep learning. Related packages include

Hi all,
i want to know some information regarding time series modelling using auto.arima package..
upon counts.. like
i have 414 rows of counts for each date .
i am forecasting using auto.arima but its not getting me nearer to true match..
although its in between hi90 and lo90 but its range is far high.
i had also tried to forecast using holt winter and fourier transformation and tbats nothing getting me nearer to true match..
kindly suggest appropriate method.
my data looks like below.


you'll find that you'll get better help here if you do the following:

  1. Create a reproducible example (reprex) of what you have tried. A reprex is code that other people can run and see exactly what you see. So the data needs to be in the example.
  2. You have a number of aspects of your question that are very specific to your data and your methods. For example, your hi90 and low90 values. We can't begin to guess why you're seeing what you're seeing without the reprex mentioned above.
  3. Tell us what you mean by "true match"? Are you holding out some data and and trying to forecast what you held out? You know that no model will exactly match reality, right?


ya true but atleast close to that by some 100 deviation approx..
yes i divided 414 dataset into test and train and traying to forecast the hold data..
90 % train and 10 % test .
how to do reprex? can you guide?


Here you go: