Open alexkokorin2 opened 6 years ago
Is M the mean or the seasonal frequencyThanks for the link to that library. It looks like a good one.
Right now the seasonal frequency is computed automatically. The TimeSeries class contains information about the period of time between observations and the Arima interface is where to set the seasonal cycle of the model. For example, the most common use case in the domain I work in is to have a time period of one month for the TimeSeries observations and a seasonal cycle of one year. The seasonal frequency is computed here
As I review this code, I think it might be better to give users the option to set the seasonal frequency explicitly. I think I definitely want to add an Arima builder as well.
Seasonal frequency.
When I run this code for input: https://pastebin.com/Y1fKKf0d I get extremely flat output (basically 300 of 15 and 420 of 16). Isn't it really strange?
TimeSeries timeSeries = TimeSeries.from(Doubles.toArray(timeseries));
ArimaOrder modelOrder = ArimaOrder.order(0, 1, 1, 0, 1, 1); // Note that intercept fitting will automatically be turned off
Arima model = Arima.model(timeSeries, modelOrder);
// Read forecast values
Forecast forecast = model.forecast(forecastSize); // To specify the alpha significance level, add it as a second argument.
I know my (p,d,q)(P,D,Q)_M (seven) coefficients, where should I put them? This library has only 6 coefficients as far as I could see.
This library has a nice sample code in README: https://github.com/Workday/timeseries-forecast