Article Preview
TopIntroduction
Consider how the aspects of food are fuzzy. Each food contains a multitude of nutrients. Many of these have been measured, but there is a great deal of variance depending on a number of factors.
Additionally, the amount of a nutrient a person should consume is not some discrete value, but mimics a dose-response curve as shown in figure 1. Insufficient Vitamin C would result in scurvy. Too much Vitamin C is toxic, but there is no crisp value immediately below which one would be healthy and above which one would be unhealthy or vice versa. The degradation of health is in direct response to the degree of intake.
Figure 1. The dose-response curve for essential elements
There is also individual vagueness such as how well each nutrient is absorbed, and sources of uncertainty in how much food an individual consumes.
The idea that nutrition should be approached in terms of continuous logic, rather than discrete, was first noted in (Uthus & Wirsam, 1996). The authors of this work found this compelling, and in their previous work (Krbez & Shaout, 2013), the literature was explored for uses of fuzzy logic in nutrition oriented systems and deployed an FLC to inform the user of nutrients to seek out or avoid based on a dose-response. This proof of concept showed the benefit of fuzzy logic in nutrition logging.
In this work, the authors take several approaches towards furthering the use of fuzzy logic in nutrition logging systems:
- •
Review past work in fuzzy nutrition systems;
- •
Explore features used in popular nutrition logging systems;
- •
Identify uncertainties that haven’t been modeled in past works;
- •
Discuss how non-singleton interval type-2 fuzzy logic may be more useful for nutrition than the type-1 fuzzy used in the previous work’s prototype (Krbez & Shaout, 2013);
- •
Implement a web application based on the most useful concepts above.
The article is organized as follows: In the next section, background information is covered. Past work on combining fuzzy logic and nutrition is reviewed. The features of non-fuzzy diet journaling software are enumerated and evaluated. An overview of Type-1 and Type-2 fuzzy logic controllers is given, including the strength of Type-2 over Type-1 in its representation of uncertainty. The differences between singleton and non-singleton inputs are evaluated. Fuzzy arithmetic is explained, especially with regard to piecewise linear fuzzy sets, which can be used to model uncertainty when accumulating input data. Modeling nutrient value uncertainty as fuzzy geometries is discussed, followed by an overview of uncertainty for nutrition labels and a standard reference database, culminating in a methodology of combining these data to extend a model of nutritional uncertainty with an IT2FS. Dietary reference intakes are explained and used to produce antecedents for an IT2FS.
In the system requirements/design section, a philosophy of uncertainty is given, and the topics covered in the background section are combined to describe a coherent diet journaling system that models all the previously described nutritional uncertainty in its calculations.
Then, in the system architecture section, the component architecture is described, including class diagrams for each managed component. The user interface for the developed system is described in the user interface section. Following that is a section comparing the system in this work and the previous work (Krbez & Shaout, 2013). Finally, the last section makes conclusions based on the results of the implemented system.