My project idea is to create clusters of typical rooms for hotel customers, indicate which rooms are similar to the one they are looking for and let them understand why they are conceptually similar.

I engineered this solution:

First I create clusters using K-means algorithm.

Then I create a 'cluster' column inside the database. Now every row has a new information, which is hopefully going to help customers to choose which rooms are (in terms of facilities, amenities, and price) similar.

Then I pass this new database (of course discretized) with the new information to my Bayesian Network algorithm, and I set the 'cluster' attribute as class attribute for my Net. Note that every information of the knowledge base is known a priori. I just need to check correlations between features when considering clusters part of them.

I'm choosing the cluster attribute to be the class because it is the only attribute describing (semantically speaking) every single room: other features can't be univocally used to understand an hotel room.

Results are encouraging, because in an initial test I found a 90% TP-rate, but I'm fearing this is because there is some major correlation between the features influencing K-means and the Net.

Mind that, even though i'm an AI student, I'm not an expert and I'm asking this question to understand if my reasoning for tackling this problem is correct. Also this is not the full project idea, I'm just conceptually stuck up to this point.

Every idea, tip, guidance will be appreciated, but please don't act like I did a crime if this is conceptually and theoretically wrong: I want to learn what to do.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.