Packages to forcast a date of sale?


#1

Hello,
I'm just starting using R studio and have utilized a text analytics package that was very helpful in the development of a word cloud and led me to think about other packages.

I have a large amount of data on new and used tractor sales and based on variables (dealer profit, options and spec, accessories, days on lot, ect) I want to forecast an estimated sale date for each particular serial number in inventory. This will be based on the historical sales data available for like models for each dealer in the data set. Just curious, is there already some type of package that will do this? Seems like something someone may have published already but so far have been unable to locate anything and think building my own at this point might be beyond my skill level. Eventually I would like to scrore each serial number in inventory (0% to 100%) based on being a good order for the dealer to have in inventory and hopefully influence future order habits.

Thanks and looking forward to any feedback!


#2

Assuming the title of the post reflects true requirement it is a problem of modelling failure time data and R got a number of packages to deal with that. Check the CRAN Task View for a range of available packages.


#3

Hey there, agricultural economist here. Sounds like an interesting model! As is often the case, this feels like a simple linear regression to me. I think your dependent variable is something like "number of days on the lot", not the date. You don't really care about the date, you care about length of time from start of sale to end of sale (I think). Then you can start making features that are predictive of duration to sell. It sounds like you have a method of picking analog sales. I'd create an index that's something like "% difference between list price and average analog sale price". So if the tractor is of a type that typically sells for $50,000 but it's listed at $55,000 then your index would be 10%.

Turning that into your 0-100 index should not be done until you've completed the above modeling. You may discover things like: 15% underpriced lawn tractors sell really quickly. You don't want a model that tells the dealer in Des Moines that they should stock only lawn tractors. That dealer is likely interested in having a diverse inventory which fits their market.

Good luck!


#4

Wow, thanks...very informative write up. Just curious...is there a package out there for this? Very new to this type of analysis so hoping to find a step by step guide.


#5

No. But there is an entire speciality called “econometrics” which focuses on regression analysis.