This is a piece of software to perform data mining (currently owned by IBM and known as SPSS Modeler) for which we’ve a number of active licences since the British Steel days. I supposed it was an early version of the type of data analysis the likes of Facebook and Google now use to analyse customer habits and preferences.
I haven’t used Clementine for a long time, mostly because my job in the last 15 years in Corus / Tata didn’t really call for it, but also because I’ve seen it used as a data extraction tool where a bespoke .NET application could of the job far easier and fully automated as well.
The one time I used it properly in an attempt to find a pattern determining the flatness of temper rolled blackplate ended in failure. I thought my sample of about a 1000 records should have been plenty – it definitely took a long time and much blood, sweat and tears to collect the data prior to automated data gathering – but when you split your data set over various reel types, two types of annealing, various gauge ranges and mechanical properties you found that many of the conclusions for a possible correlation were based on ridiculously small sample sizes, and therefore any conclusions were not worth a sack of beans.
It definitely cooled my ardour for this type of investigation, and fortunately my subsequent job content no longer called for its use. Maybe that’s all for the better, since I’m not aware that all those years of using Clementine in Operational Research have really led to any new insights – at best they merely confirmed what a knowledgeable practitioner might have suspected prior to the analysis. Our investigation into the issue of PM10s (see Pollution) was a case in point, snce it merely onfirmed the prime importance of wind direction.
There was also some misuse of Clementine in trying to do some jobs that were better performed by a bespoke .NET application. One example was the extraction of screen sales records, which meant someone had to kick off the Clementine job once a week, capture the resulting records in an Excel spreadsheet and snd this spreadsheet to the interested parties. Exactly the type of labour intensive job that I was trying to get rid off – so that’s what I did : I used the logic for collecting the relevant screen sales records to build an extraction application, populated an SQL table, and built a web page to display the results. Meaning that from that moment onwards no further manual input was required to retrieve and analyse the data.