1
$\begingroup$

I'm currently building a model to predict internet auction sale prices of products in a marketplace. There are a lot of instances where a product goes for multiple prices but it's basically the same item. I have features such as category of item (clothes, kitchen, bathroom, etc.) color, weight, height, US state the item is currently in, condition of the item (good, bad, ok), and a number of other features. I have historical data dating back five years which allows me to see how much items have gone for in the past.

There are occasions when for example a bicycle might be basically the same item but there's a price variability of 100 dollars between the two items. Time of day, month, auction length help some bit but I can't seem to account for the seemingly random way that prices go up and down. One bicycle might sell for 150 dollars and the other bicycle sells for 250 dollars.

I am using Python to make the model.

$\endgroup$
  • $\begingroup$ Welcome to the site. Until your final paragraph, I was following along and thinking about solutions. But what has random walk to do with this? $\endgroup$ – Peter Flom Apr 17 at 11:44
  • $\begingroup$ @PeterFlom I don't honestly know much about random walk but from what i've read about stock prices, random walk assumes that stock prices can take an unpredictable path compared to previous trends. So what I was thinking random walk could be incorporated is if I have ten bikes and they went for ten different prices on a day, the random walk would be able to randomly assign integers to increase or decrease a price prediction. Again, this is my limited understanding of it and it's probably off base. It's probably not that relevant and I can remove it if it distorts the question too much. $\endgroup$ – DS_Misc Apr 17 at 15:29
  • $\begingroup$ I would delete it. $\endgroup$ – Peter Flom Apr 18 at 11:51
  • 1
    $\begingroup$ Thank you @PeterFlom. That part has been removed. $\endgroup$ – DS_Misc Apr 18 at 16:02
1
$\begingroup$

This doesn't really seem like it should pose a problem. If you run a regression with "price" as the dependent variable, then there will be variation in price, even given all the independent variables.

But that's common; I'd say it's almost universal.

$\endgroup$
  • 1
    $\begingroup$ Thank you Peter! $\endgroup$ – DS_Misc Apr 18 at 21:43

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.