0
$\begingroup$

Here, I want to predict dealers(about 600) whether pick up the goods(about 30) next month. As you see, there are about 18,000 possibilities and it's difficult to predict.

By the way, now I have nearly three years'data(2017,2018 and 2019), and now I just use the common object in the last several months as the prediction.

For example, if dealer A purchases the goods item 1 in September, October and November, then I predict he will purchases item 1 in December. However, this is not exactly accurate, and the precision is about 0.6 while the recall is about 0.3.

So any other way to predict it?

$\endgroup$
3
  • 1
    $\begingroup$ What algorithm are you currently using to predict that? If you are using regression, how are you performing it? $\endgroup$ – Harry Stuart Dec 18 '19 at 6:29
  • $\begingroup$ Now I don't use any algorithm, just use some rules to get prediction. $\endgroup$ – rosefun Dec 18 '19 at 7:31
  • $\begingroup$ My suggestion would be to look into logistic regression. It seems quite appropriate for your task. $\endgroup$ – Harry Stuart Dec 18 '19 at 7:43
1
$\begingroup$

Sure.

Question you should answer is it multiclass or mutlilabel.

In any case I would really advise you to take a look at this list of similiar problems/competitions. Since your question is open-ended I could write 10 pages of approaches. Yours best bet is to invest some time, find similiar problem and look what other people did and dig it up.

$\endgroup$
1
  • $\begingroup$ Thanks, actually I think my problem is more difficult, because it's similar to time series prediction, and there are not many samples (only three years' data). $\endgroup$ – rosefun Dec 19 '19 at 2:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.