Neural Network : How to use it to guess a wine price

alois_borgognon

Hello,

I need to guess the price of different Bordeaux Wines. For this, I will use different data such as Parker Note, Vintage Note, Exportation and Export Figures and I think to set-up a neural network.

However, it's my first time and I don't know how to optimize it with the weight, or the hidden layers, for example. Do you have any ideas to set up it well ?

Furthermore, I need to check with a performance test. Which one is the best for my process ?

Thanks a lot for your future answer.

lionelderkrikor

Hi @alois_borgognon,

Difficult to answer.
You can begin by submitting your data to AutoModel...(Maybe the neural network is not the best model for your data.)
Anyway, if you want absolutly use a neural network model,  in AutoModel there is a model called "Deep Learning" you can open the process and then play with the parameters of this model.

Regards,

Lionel

lionelderkrikor

Happy new year 2020, Aloïs !!

If you are a real newbie, I encourage you to begin 2020 by seeing the videos of the RapidMiner Academy which
explains the methodology and the basic concepts in data science : 
https://academy.rapidminer.com/

To know if your process is efficient (or not) , you are defining a performance metrics which will quantify how well you model fit to your data. In your case (regression model), this performance metrics can be RMSE.
Then you will build some models (decision tree, neural networks and other models) and calculate the performance metrics for each of these models and then you will choose the one which performs best.

I insist, if you are a newbie, : you can submit your data to AutoModel and RapidMiner will perform automatically all the tasks discussed above for you. RapidMiner will build some models with your data, calculate and present the performances of each of the models. You just have to pick the best one..

For your question about the best neural network architecture for your use case, I will need the help of my friends, @varunm1 and @pschlunder , the deep learning gurus of the community ...
I advice you to share your data in order we can give you a more relevant help ! 

Regards,

Lionel 

lionelderkrikor

Hi @alois_borgognon,

Difficult to answer.
You can begin by submitting your data to AutoModel...(Maybe the neural network is not the best model for your data.)
Anyway, if you want absolutly use a neural network model,  in AutoModel there is a model called "Deep Learning" you can open the process and then play with the parameters of this model.

Regards,

Lionel

alois_borgognon

lionelderkrikor

Thank you for your comment !

I have some another questions. Which kind of processes you suggest to guess the price ? I think also to use a regression tree. What do you think about ?

Furthermore, about the deep learning, do you have any advices to set it up   ?  How may I know if my process is efficient or not ?

Sorry for this disturbing, I'm a real newbie.

Have a good year too

Regards,

Aloïs

lionelderkrikor

Happy new year 2020, Aloïs !!

If you are a real newbie, I encourage you to begin 2020 by seeing the videos of the RapidMiner Academy which
explains the methodology and the basic concepts in data science : 
https://academy.rapidminer.com/

To know if your process is efficient (or not) , you are defining a performance metrics which will quantify how well you model fit to your data. In your case (regression model), this performance metrics can be RMSE.
Then you will build some models (decision tree, neural networks and other models) and calculate the performance metrics for each of these models and then you will choose the one which performs best.

I insist, if you are a newbie, : you can submit your data to AutoModel and RapidMiner will perform automatically all the tasks discussed above for you. RapidMiner will build some models with your data, calculate and present the performances of each of the models. You just have to pick the best one..

For your question about the best neural network architecture for your use case, I will need the help of my friends, @varunm1 and @pschlunder , the deep learning gurus of the community ...
I advice you to share your data in order we can give you a more relevant help ! 

Regards,

Lionel 

alois_borgognon

Hello LIonel !

Here is my database. In a nutshell, I would like to guess, with efficiency, the price of each wine castle for any vintage with the help of the ranking (it hasn't changed since 1973), the grade of each castle for each vintage from wine global score, the vintage grade (Latour + idealwine) and the exportation number (USA, China & Hong Kong).

Thank you so much for your help !

Regards,

Aloïs

alois_borgognon

Hi Everyone !

My apologize. I Updated and optimized my database for rapidminer. I think it will be more efficient, like this, for the neural network.

Thank you for your help ! You're amazing !

Thanks a lot again,

Aloïs

alois_borgognon

Again sorry, I updated my database...

alois_borgognon

@lionelderkrikor, @varunm1&nbsp,  @pschlunder,

How may I set up a neural network with this database ?

@lionelderkrikor : I tried the AutoModel and here are results : 

RMSE :

• Generalized Linear Model : 138.918+/-79.683
• Deep Learning : 123.879 +/- 70.732
• Decision Tree : 91.116 +/- 57.742
• Random Forest : 99.553 +/- 60.853
• Gradient Boosted Trees : 100.984 +/- 61.276
• Support Vector Machine : 65.184 +/- 32.665

I think I have a real issue with my database. How May I improve it ?

Thanks a lot for your helps. You're really incredible !

Regards,

Aloïs