package which provides a accurate forecast higher than 50%

My challenge is the following:
I have an univariate time series.
I need to have the correct R code which has at least 51 % or higher accuracy to forecast the alpha,beta and gamma numbers.(ABG)

The below 2 sets of code provide each the ABG .But nut accurate enough.

sales <- ts(sales, frequency=3, start=c(2019,1))
predict(ts(sales))

sales <- auto.arima(sales)
plot(forecast(sales,h=20))
forecast(sales)

<sup>Created on 2020-06-25 by the [reprex package](https://reprex.tidyverse.org) (v0.3.0)</sup>

Is there a package which provides a  accurate forecast higher than 50% ?

I don't understand this request. model accuracy is largely determined by the underlying system which is attempted to be modelled. If I have an entirely predictable system, then I can achieve 100% accuracy, if I have a perfectly random system, I can't expect any accuracy. most real world scenarious are somewhere between, I would question why one would expect business sales to be more than 50% predictable...

2 Likes

Without more information it is impossible to help you improve this time series model. You also seem to be asking for something that is not realistic...A package alone will not improve your model's accuracy. Time series models are tools and if you want to use those tools you need to understand them. A package can't just "solve" a time series problem for you.

A nicer calculator won't make you better at math - but it can help. This is kind of the same situation.

Here is a book on how to do time series in R: https://www.amazon.com/Practical-Time-Forecasting-Hands-Analytics/dp/0997847913

There are also several youtube courses available for time series, including a full MIT course on the subject.

1 Like

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.