Introduction

The purpose of this project was to create a case based reasoning system that helps select holidays for people. The idea is that someone enters into the system their ideal holiday, and the system tries to find the most similar available holiday from its case base.

The important difference between this and standard database retrieval is that in this system, certain compromises must be made. For example, if a five-star hotel is requested along with a request for a relatively low-priced holiday, a compromise such as increasing the price or downgrading the accommodation must be made. This system makes those compromises and returns the best matching holidays, even if they are quite different from what was originally requested.

We are supplied with 1024 cases for our case base. An example case is as follows:
 

case	Journey11
JourneyCode:	11
HolidayType:	City
Price:	1978
NumberOfPersons:	2
Region:	Cairo
Transportation:	Plane
Duration:	7
Season:	April
Accommodation:	ThreeStars
Hotel:	Hotel Victoria Cairo.

 
For each dimension listed above, we define a similarity function. For example we may say that May is 80% similar to April, 60% similar to June, and 0% similar to November. We can then add up all the individual similarities, weighting the more important ones such as price, to get an overall similarity between two cases. We can then find the most similar cases to a new case, and return these to the user. Because we do not adapt the returned cases at all, this is really a "Cased Based Retrieval" system.

Click the links below to see how this was achieved and to see it in action.