I am working for a startup/app company that is trying to assign a dollar value to virtually every apartment in New York City, based on several unique characteristics of our users. What we need is a model that predicts the unique value to YOU (the user) of each apartment, based on factors like where you work (commute time) and the locations of other places that you frequently visit (the address of a girlfriend/boyfriend for example, or perhaps an entertainment venue that you frequently visit).
The model’s inputs might include:
1. How much you (the user) value your time. This can be represented in $ per hour.
2. Apartment address
3. Apartment’s advertised monthly rental rate
4. Workplace address
5. Location #3 address
6. Location #4 address
8. Location #n address
Also keep in mind, we already have built a program that takes physical addresses and calculates the travel time between them in real-time. So if you had just two locations (home and work), we could calculate the average morning commute time and the average evening commute time from this program. Thus, a simple base case formula might look something like this:
V = P – 20*(T1*W + T2*W) – C
Where V is the value of this particular apartment to YOU, represented in dollars per month,
P is this apartment’s advertised monthly rent (in dollars per month),
20*(T1*W + T2*W) – C represents the value lost from having to commute, as opposed to just living at work,
20 is the number of days per month most people work,
T1 is the morning commute time,
T2 is the evening commute time,
W is your approximate hourly wage (or how much you value your time),
and C is the cost of commuting (e.g. a train pass).
As you can see however, as soon as Location #3 is added, the model becomes a bit more complex.
For this project, I need a skilled and experienced mathematician to design an algorithm around this more complex scenario. Also the outputs have to be realistic. For example, no apartment can have a value that is negative…you just would never rent those!
For the scope of the project, let’s keep it focused mainly on commute time for now, since the number of inputs could otherwise be virtually limitless. Also, please definitely provide graphs and charts, if you are capable, to display the output in various scenarios. I am a very visual person, so this would help me understand your solution. The ultimate solution can be in plain English or using a computer program, but must be easily understandable by me and my software team.
Please contact me (Darrick) if you are interested in helping out or have any questions. Let’s make history together with a new product that will help serve nearly every renter in New York City!
P.S. Also let me know if you see an easy solution right away. For all I know, this could be a common Ops Research problem that I just have not studied yet.
Hello everyone! Just to clarify, I am mainly interested in someone to help with the algorithm and/or mathematical formula for this computation. Not the actual programming (we already have programmers to implement it in-house). I also do not currently have a data set to provide. We are still trying to acquire the data. We simply need a way to model the problem I have described above, in the case where there are more than 2 locations. For now we are strictly looking at commute times as the inputs. Obviously an apartment valuation model could include thousands of other variables like floor/story, square footage, year of construction, etc. etc. But we do not want to include those until later on, once we have the commute time part figured out.
Thanks! I have been reading your applications and hope to pick someone by this weekend at latest.