R package to derive probability that any two food items are the same
FoodC5 is an R package providing an interface to a pre-trained C5.0 classification model. This C5.0 tree enables to predict whether any two food items are the same. It requires the following input parameters:
- fuzzy score: similary between the two food names
- information on how much the two items differ in energy content
The fuzzy score can be computed with with the Foodmapping R package. The difference in energy content can be expressed as 100 * abs( itemA_energy_content - itemB_energy_content ) / itemA_energy_content.
To install, run the following commands in R:
install.packages("devtools")
devtools::install_github("armandvalsesia/FoodC5", build_vignettes = TRUE)require("FoodC5")
# load example dataset
data(food_score)
# Compute probability based on a comparison using the English-translated food names
preds <- c5_one_language(food_score, 'ENG_SCORE', 'EDIFF')
# Compute probability based on a comparison using the original food names
# this model is trained for the following languages: English, Dutch, Danish, and a modest subset for Greek, Spanish and Bulgarian.
preds <- c5_one_language(food_score, 'ENG_SCORE', 'EDIFF', c5_object = read_c50_ori())
This software uses the GPL v3 license, see LICENSE. Authors and copyright are provided in DESCRIPTION.
Big thanks to Hélène Ruffieux and Radu Popescu for help with CI.