Message Boards » Website Suggestions/Feedback

TOPIC: Suggestions for Better Data Mining

 
June 21, 2012 13:14
I have a suggestion to help MFP members better utilize past logs. It involves expanding the "Reports" tab to dig deeper into the data. I would suggest looking to other data analysis websites for ideas. Those that first come to my mind are Mint.com, Facebook Insights, Google Analytics, and Hootsuite.com. All of these sites have highly user-friendly means of presenting information mined from very broad and deep datasets.

I have one specific idea for a report based on Mint.com's interface: a Report of "Top 10 Calorie Sources" (or "Top 10 Fat Sources," "Top 10 Carb Sources," or any other dimension of nutrition). It would be very helpful to know which foods comprise the bulk of calories. This information may not be obvious just by entering daily logs, but rather is more clear when observed over time. For example, a user may not realize the two tablespoons of peanut butter they eat every couple of days accounts for a disproportionate share of their monthly caloric intake. This could help users hone in on problem areas to improve their diets--either by cutting portions or replacing with a more healthful alternative.

I realize there are some difficulties associated with generating this kind of Report. With the current database, it seems a user would have to input the exact same food entry every time in order for that food to show up as a repeated calorie source. Since there are multiple entries for pretty much every food imaginable, this would reduce the usefulness of such a tool for many users.

Alternatively, it would require substantial additional metadata--for example, categorizing all the different peanut butter entries as "peanut butter," all the different brands and types of rice as "rice," and all the different butters, oils, and shortenings as "cooking fats," just to name a few examples. But I can imagine that a well-designed search algorithm or something like that may be able to automate some categories, like however Mint.com categorizes charges into categories (groceries, gas, restaurants, etc). I don't know if that level of metadata is available or even within the purview of future enhancements to MFP's database, but I think it expand our ability to leverage the swaths of data we are generating.

Reply

Message Boards » Website Suggestions/Feedback

Posts by members, moderators and admins should not be considered medical advice and no guarantee is made against accuracy.