I have to predict rank(1~100) of online sales products.
There are 'keyword' and 'date' columns, so each keyword and each date, there are 1~100 rank.
And for X, there are 20 variables(review number, price, points, etc).
There are more than 100 different keywords. And X values for each keyword have different scale.
Which model would be appropriate?
I tried multiclass classification model, one model by one keyword. And for evaluation metric, I used spearman's correlation. Score was 0.5 ~ 0.9(it varies by keyword).
Any suggestion would be big help.