case-base archive

ai-cbr is establishing an archive of case-bases for use by researchers and developers. The archive will accept any case-bases in any format. These can be in a proprietary tool format (e.g. CBR Express, ReMind, CBR-Works, ReCAll or ESTEEM formats) or in any tool neutral format such as CSV, ASCII, or database and spreadsheet formats. ai-cbr will place links from this page to the case-bases on your server or you may email the case-base to ai-cbr and we will host them on our servers.

There are many data sets in the UCI ML Repository that are suitable for CBR research. The main distinction between data sets held there and case-bases here is that the ML data sets have a purpose (i.e., a data set may contain categories that an ML algorithm could discover). Case-bases do not necessarily have a purpose in the ML sense.

MLNET also maintains a list of links to ML data sets.

 'ArchiveLady' PReill

Please note that these case-bases are offered entirely unsupported. ai-cbr cannot help you with any problems you might encounter with them.

travel agents case-base

Approximately 1000 cases each with 11 attributes describing different holiday/hotel destinations. Because of the practical size and wide range of feature types this case-base has almost become a de facto standard within the CBR community. If you've invented a "better" retrieval algorithm you should test it against this case since you will then be able to compare your results with many others.

 

office buildings case-base

(ReMind version 1.0 format) Approximately 80 cases each with approximately 40 features. The case-base includes very extensive adaptation formulae.

 

pest control case-base

Used by CARMA a system of Karl Branting's described in ICCBR98

 

music case-base

These are the three music pieces used in Luis Macedo's system INSPIRER/SICOM to compose new music pieces described in ICCBR97 and ECAI98. The composer is Carlos Seixas, a XXVIIth century Portuguese composer. The format of the files is Prolog.

 

soil case-base

Soil type is predicted based on the concept that soil type rests in certain landforms. If we know the landform then maybe from the past experiences we can predict the soil type for a new plot. The zip file contains  MS Excel and CSV formats.

 

design & estimating case-base

This is a collection of dbf files from the NIRMANI case-based design and estimating system. The system uses a hierarchical case representation which explains the number and nature of the dbf files.

 

food case-base

This is a simple example using a nearest neighbour algorithm to select food based on a persons preferences. It is implemented within a Lotus Approach database.