Sunday, March 29, 2009

The Amazing Computer Science Diet

Ingredients:
  • A weight tracking site like FitDay. People enter some information about themselves, what they eat every day, and how much they weight. The site lets them track the number of calories they eat along with their weight. Currently, these sites don't make recommendations about what to eat.

  • Data mining algorithms.

Preparation Instructions:

Once you have over 5 million users on your weight tracking site, perform statistics to find out what actually makes people lose weight – e.g., “by eating one more cup of yogurt every day, you can lose 1lb per month.” As opposed to other diets that are pulled out of somebody’s behind, this one will be based on millions of data points.

Start recommending what to eat.

Alternative (Advanced) Preparation Instructions:

Use collaborative filtering to determine what is the best diet for each individual. By looking at people who have similar profiles to each other (they weight the same, like to eat the same things, etc.), it may be possible to design a diet that works for you personally: “that person who is very similar to you lost weight by doing X.”

28 comments:

  1. I think something similar is supposed to make its way into iphones soon...

    ReplyDelete
  2. add social networking to said site to attract more people. The VCs will like that :) Charge for more accurate recommondations

    ReplyDelete
  3. Brilliant idea. Can i steal it?

    ReplyDelete
  4. Cute. But I suspect a more effective approach would be the one Stephen King described in Quitters, Inc.. As far as I can tell, it's not that hard to find an appropriate diet; the challenge is in the execution.

    Still, no reason not to get rich quick by offering hope to those who have cash but lack discipline.

    ReplyDelete
  5. NeedInstructions.com – Instructions Manuals & How To’s On Just About Everything Out There!

    ReplyDelete
  6. One problem would be exercise: to be able to nail down how much of the weight loss is attributable to diet, you'd need to know how much the users exercise. I'm not sure that this is entered into the dieting websites.

    ReplyDelete
  7. Averaging across many users may remove that problem.

    ReplyDelete
  8. For the other website ideas you have suggested so far, there was always an incentive for a person to give as much information as possible (i.e more high-quality porn of that type).

    However, for this particular website, the incentive for the user to give accurate information is unclear. How would you ensure that every person gives the most honest answer especially when their diet experience wasn't so great?

    ReplyDelete
  9. The incentive here is to track your own weight and caloric intake. Why would you like to the site? I know a couple of people who use these sites, and they try very hard to tell the truth.

    ReplyDelete
  10. Too many variables means practically useless.
    Just recalling what you ate could put a dent in your intake.

    ReplyDelete
  11. I disagree. I think you can find trends when you have so much data.

    ReplyDelete
  12. Suppose I'm a user of this site and I learn that some method has had quite a bit of success so I try it. However, instead, I have side effects and gain weight as a result. As a user that lacks patience (which is the case for alot of people that visit this site), I lose trust in this site and look for some other method from a different site. I think losing these types of users is fine. However, his not-so-positive experience of this diet is still valuable.

    Therefore, there may be a imbalance of positive and negative experience of diet techniques.

    ReplyDelete
  13. I am sure Luis you are not taking this idea seriously yourself because if you were than you wont broadcast this to the public.

    ReplyDelete
  14. Sure I take it seriously. I just have many other things to do :)

    ReplyDelete
  15. If you can factor in genetics, history, environment, preferences, "eating one more cup of yogurt every day, you can lose 1lb per month"
    into 95% non-contradictory data points, then why not the stock market? heck, why not phrenology?

    ReplyDelete
  16. I really like this site, and I think it will be useful for my personal weight loss.

    However, I think the idea is pretty bad. Sure, there's a lot of data, but people submit data voluntarily. There is bound to be huge amounts of bias.I would take any recommendation made by such a site with a grain of salt.

    Instead, why not perform a real statistical study or experiment on say, a thousand randomly sampled people, and use that to make recommendations?

    Oh wait, many organizations already do this.

    ReplyDelete
  17. I claim that a larger sample of potentially biased data may be better than a smaller sample of mostly unbiased data.

    ReplyDelete
  18. I love the idea.
    Any one reading this post wanna form a team?

    ReplyDelete
  19. Oh, Yeah?
    Check the Netflix Prize for a "large sample of potentially biased data" before you jump!
    (LOL)

    ReplyDelete
  20. I'm liking your get rich quick schemes... that said, I think that the flaw with this one is that what you're suggesting isn't that different from current practice, just with finer grained information.

    In particular, you'd basically be constructing a large scale cohort study like the Nurses' Health Study, but you'd be polling constantly. The constant polling wouldn't help the fact that many of the patterns you would be looking for would be long term, however. (Also, there's Bonferroni's principle re: the yogurt example.)

    ReplyDelete
  21. 5 FACTOR DIET - everything you need to lose weight fast!

    ReplyDelete
  22. I forget to eatwhen l am on the computer so have no need to monitor it, ha ha.

    ReplyDelete
  23. When l am in the limo, and waiting on a client the first thing l do is switch on the lapto and start to munch, this will good idea for me.

    ReplyDelete
  24. Great tips, I think a lot of people hit that too.

    ReplyDelete
  25. What a great info, it's so useful for me. thanks.

    regards,
    Anthony

    ReplyDelete
  26. why not perform a real statistical study or experiment on say, a thousand randomly sampled people, and use that to make recommendations?

    ReplyDelete
  27. Good post. But still want some more information on computer check supportonclick

    ReplyDelete