I have been working on this idea that I haven't found enough devlopers for. The development is thus quite slow and of course, I haven't been able to get any investments either.
But here is the idea anyways. When you visit a new city (for purely tourist purposes) you have little idea of where your favorite places can be. Of course, it takes time to find out what your favorites are. But what if a friend could tell you where you should go, a friend who has known you for years and understands what kind of places you would be really interested in. If you are an art-lover like me, the friend would've told you about the museums and ongoing exhibition in the city, or may be a journey to the house of a famous architect outside the city followed by relaxing in a neighborhood bar where mostly artists hang out.
Well the idea is to have a website do that for you. networking websites like facebook, orkut already know a lot about you. They know what kind of a person are you and what kind of people you would share your interests with. Even if they don't a little survey can tell a lot about your personality, objectively.
I came up with a clustering algorithm, in order to organize people in different categories. There are no rankings of parameters in the whole algorithms. Everything is based on the distance from individuals. In other words, the algorithm doesn't try to calculate how much you like something but instead it only purports how far are you from a museum lover or a party animal (a distance metric based on the feature vector in the clustering).
One the algorithm discovers your personality - it suggests you what places you should go, fetches schedules from the museum sites and tells you what train to take. If the trains are not running, it tells you what other means you can take. It calculates fare as much as you can and adjusts the costs that you have already given to you. It even takes the weather forecasts and asks you not to go to the park when its raining... and a lot more.
Tell me which real friend would do that much for you..
The opposing argument is one that favors uncertainty. Of course, there is fun in serendipity. Nothing takes away the joy of discovering things that you hadn't planned. But i mean, you can't plan for accidental discoveries and that doesn't make planning worthless.
2 comments:
hey,
is this not a typical usecase for a recommender system? have you thought of any insights into the problem that would make it fundamentally different?
bala.
ps: we haven't spoken to each other in a long time. i am kind of busy with work in an unpleasant way, but i want to catch up with you sometime soon.
right! its nothing more than a standard recommendation system - it was inspired by amazon and netflix (way back!) .. I just didn't do much development.
The key is to integrate the data and optimize with constraints (like weather, subway schedule, preference of one medium over the other). I think can now write a small app for iPhone. I am sure there is something like this already being developed somewhere. But I thought I had to something with this idea :).
PS: Whats the best number to reach you?Would it be the 12***51 number? Is it OK to call around 10-11pm IST?
Post a Comment