Week 5 — Effectivity of Ingredients
After a long time of crawling a data-set from allrepice.com, we have at a stage which is to create a concrete algorithm and cleaning noise data in a dataset in order to solve this problem during the recent week.
As we mentioned in the progress report, we came up with a solution which leans on the idea of TF-IDF. Before telling much, let us see the illustration of our algorithm.
N shows all ingredients in the data. U refers the ingredients that user has eaten in a meal but did not like the taste. C depicts the ingredients which user has eaten in a meal and did like the taste.
It seems that this algorithm is not that complicated from outside and easy to understand. However, when it comes to some math equations and formulas regarding to have a meaningful results, it is greatly complicated since there are many factors that influence correctness of the algorithm.
As it could be inferred from text, we presume to use Naive Bayes theorem. As the input we think about using TF- IDF data of the ingredients existing in the whole recipes.
On the other hand, the data-set which we crawled, has many noise data. Thus, it takes much time to clean them in a reasonable manner. However, it is a matter of time to reach the perfect result.
Now, you have better understanding about what we do this week, see you in the next progressive week !! ! ! !
Reference
