Real time domain specific algorithm for filtering foods by a multi-dimensional constraint matrix containing a health profile

Kusper Gábor Dr. <>
Eszterházy Károly Főiskola

Márien Szabolcs <>
Wit-Sys Zrt.

Kovács Emőd Dr. <>
Eszterházy Károly Főiskola

Kovács László Dr <>
Miskolci Egyetem

The aim of the eFilter project is to setup an information system, which can filter the list of foods by personal health data. The food can be a very basic one, like flour, or a complex one, like macaroni with cheese. The personal health data is stored a so called health profile, which is a personal health record. This contains all food sensibilities, allergies and diets of the person and any other health problem which constraints eating. This information is stored in a numeric format. We do not store, that the user has hazelnut allergy, but we store that the daily eatable hazelnut quantity is between 0.0 and 0.0. So we story constraints on food ingredient quantities. In this manner the personal health profile is a multi-dimensional constraint matrix. Based on this, we have to decide as quick as possible, whether a food is eatable or not. For this we have to know the ingredients of each food. We describe in an other paper, how to collect and validate this information. In this article we describe the algorithms which use the constraint matrix. From these the most advanced algorithm uses an indexing strategy, which allows to decide in real time, whether a food is eatable or not. For this we index the food database based on its ingredients. To depict this method let us take the quantity of hazelnut. With the first index we index those foods which do not contain hazelnut. The second one indexes those ones which contains one milligram or less hazelnut. The n. index indexes those ones which contain 2^(n-1) mg or less hazelnut. In this way with 20 indexes we can index the quantity upto 0,5. So this indexing strategy allows very fine distinction on the mg level, and a very coarse one on the kg level. These indexes can be used to filter the food database very effectively without using float number mathematics.