1
$\begingroup$

I have a dataset with features like Cylinders, Transmission, ExteriorColor, OdometerReading, StockType(New/Old), MakeModel, ModelYear, SoldYearMonth, CountSoldByMonth, Price, TopSelling.

Target column is 1 if a particular MakeModel is the top 5 selling vehicle in a month. So, given 60 months data I have to predict the next months top selling vehicles. I also have to predict the price go the vehicle. So I guess, I need to do both Classification and Regression .

MSRP   Cylinders   Transmission   ExteriorColor   OdometerReading   StockType     MakeModel            SoldYearMonth   CountSoldByMonth   Price   TopSelling
42098      8            MYC            Black           70348           Used    Chevrolet Tahoe            5/1/2019           10           39998     0
26390      4            MNK            Black            10             New      Buick Encore              5/1/2019           31           26390     1

and so on.. MSRP and Price are not always same.

How do I approach this problem?

$\endgroup$
1
$\begingroup$

You can also consider a multi-output model using the Keras functional API (see "Deep Learning with R", Ch. 7, p. 224). Here are the respective Keras docs. There are also some tutorials online. I'm not sure if a NN is suitable for your problem, but this would be an option to consider. The advantage is that your model can also take correlation between the features and the two outputs into consideration while learning (which may give better results).

$\endgroup$
0
$\begingroup$

Though I'm not sure what is MSRP, but still for such type of problems where you are expecting two different behavior, I suggest to go for two different model trainings,

  1. For classification
  2. For regression

So you have to train two different models one for price and another for top selling model using the same features set.

Note: You can use price as a feature(along with other) for classification problem and top selling values as a feature for regression problem.

$\endgroup$
  • $\begingroup$ MSRP is the manufacturer's suggested retail price, or how much the maker thinks it's worth. I would think it'd be an important feature for their model. $\endgroup$ – HS-nebula Jun 28 at 19:13

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.