Welcome to my wine project! I hope you will appreciate the journey! For a long time, wine tasters have been seen as the only trustworthy source
for giving a reliable opinion on the quality of wine. More recently, Vivino has come into the picture: it relies on community-based ratings.
My company "Weeno" (aka We Know) is a newcomer into the field of wine tasting and has to define business problems to make an effective entry into this market.
I have defined 2 business goals: one for the short-term and the other for the long-term:
- Our short-term business problem is of customer acquisition. In order to achieve this goal, a crucial part has to be how reliable we are; because let's be real, nowadays community-based platforms are taking a lot of markets by storm. This is why to be able to compete with them, we must be able to convince our consumers that our results are the most accurate possible. We will use regression to answer this first business problem by predicting the quality of wine given a set of inputs (nothing to crazy!). Our focus here must be on lowing the error rates we make as much as possible (you will see later which error metrics I chose).
- Our long-term task will be that of customer retention. Indeed, it is one thing to give accurate predictions, but what if we go one step further and predict whether the user will like the wine or not. This is achieved through a classification problem that you can find in the other file. Through this regression problem, we are going to predict whether a wine is "good" or "bad" and this will allow us to keep our customers loyal because they would think that we are giving them these predictions based on their preferences, when in reality it is all pre-programmed. However, this must remain a secret between us.
Eventually, when my business expands and my reach increases significantly I will be able to ask users for their ratings and eventually not only display the algorithm's prediction but also what others think. Now you must be thinking: but isn't the community part exactly like Vivino's model, and to that I say "YES". We are taking customers from Vivino, increasing customer loyalty and then using their own model in an indirect way! However, this is too much of a long-term thinking and not in the scope of this project.
As for the sequence of this project:
- Start by opening final_project_DA4B.R, or proj_1_reg.Rmd (they are the same files but in different formats: I leave the choice to you!): this contains the regression problem
- Then open fin_proj_classification.R, or Proj_2_class.rmd (again, same files but different formats!)
Just to spoil you, I will also provide you with compiled html versions of both taks: ENJOY!
I hope you enjoy this project as much as I did doing it!
In case there is any issue, contact me at marcviolides@yahoo.com