Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
I'am trying to create a regression based prediction (like booking website): predict the number of clicks for each hotel.
I have to generate a .csv file containing two columns:
hotel_idpredicted_number_of_clicks for all hotel_ids.My first question question is: should I put the hotel_id as feature in the predictive model? I think that I must to drop it... right?
Second question is: how can I write in the csv file only this 2 columns if I drop hotel_id from the model features?
Trying to classify using hotel ID is the same as trying to determine if a student is going to perform well on a test based on their last name. You should get additional things, like number of rooms, amenities, location, staff, etc. Those are informative features that you can use in a classifier
When training the model you don't need to use the hotel id. The model needs to learn from examples. It only needs feature values and number of clicks, so it can learn the relationship between these.
Once you've trained your model, you can use it for unseen examples. These would be hotels for which you have both an id and a set of feature values. Your model should take the id and the feature values as input, but it should only use the feature values for the prediction. The id should just be kept on the side, so it can print it together with the prediction to the output csv file.
I hope this helps!
Supervised learning should try to 'understand' what makes a hotel to have more clicks than other. As a consequence learning tries to define which are the characteristics of some given hotels which make them attractive or not. So it uses some kind of similarities, because it is supposed that similar hotels behaves in a similar way.
Now if you restrict the similarity to identity than you learn nothing new because hotels are unique. In fact such kind of learner exists and is called Rote learner, and it consists of one-to-one mapping from inputs to outputs. It is also called memoisation. And this happens if you will add hotel_id in the features. However I think you hope to use that to predict the number of clicks for new hotels (which does have a different hotel_id than any from training set).
On the other hand, in order to use hotel_id to store prediction you only have to save a copy of the original data set. At learning time you have a train data set from which you remove hotel_id, and use that for learning.
At prediction time you make a copy of the data set for later use. From the original data set remove order_id, use that for prediction and get the results. Now the predicted results have the same order of instances as the copied data set. This happens for sure in python (scikit learn), java (weka), R. In fact I am not aware of a system which does not preserve positions.
Now using positions from the copy of the original and prediction you can associate each hotel_id to each prediction with no problem.
The hotel_id should not be a feature.
Let's see if I understand you correctly.
At testing time you give your model a whole set of feature values for a particular hotel you are interested in. This hotel has an id, which is known to you.
Your model should be able to take both an id and a set of feature values as input, so it can print both to an output file.
Does that answer your question?
