Various analysis and prediction techniques were corroborated in the paper to extract knowledge from a bi-dimensional industrial dataset. It was intended to consider the least human intervention possible in all, the selection of the system predictors and dependant variables and of the data records that hold valuable information on the process. In that purpose, no prior experience was used and only results that could not be subjectively interpreted were elected (however, extracted correlations between parameters could be confirmed on an experiential base). As the paper addresses [a large amount of] practical issues, the theoretical part was omitted, though sometimes referred. This is also the case for the way several software packages mentioned at the end of the paper were used on this particular real-life analysis and prediction task. Techniques were manually iterated within a heuristic approach. In this paper we deal with the analysis of the database and the selection of the pertinent process parameters to be used in the second part of the article.

 

 

 

 

Back to Contents