Our algorithm is able to find correctly 57% of the
relations defined between th
e two databases, and ranks
the expected product in the top 3 for 83% of the POS items. It is also important to stress that only 0,8% of the
POS items are not related to the expected items at all.
For results evaluation, we took into account not only the number of successful matches, but also the
accuracy in terms of the chemical and nutritional values. We consider more important a lower deviation
between the expected nutritional values and the nutritional values found than to have a higher success rate in
terms of matches