End-2-End Predictive Modeling Process (ModelOps/MLOps)

OLAP is a precise method belonging to Data mining.  Data mining is a broader term covering a number of methods (one of them is OLAP). OLAP can be run: in software, you can ask for OLAP in a similar way you ask for correlations, regressions, and other statistics. OLAP is in a software menu (or it is a function in a language) (some packages will use “cube” or “pivot” keywords to get the same output). Instead, you cannot ask for a Data Mining execution because there is no output associated with “Data mining”, you have to invoke a method belonging to Data mining to get an output such as regression trees, association rules, clusters, support vector machines: each one has its execution in a software. In a book on Data mining, you can find a number of methods, OLAP is one of them. For instance, in the book “Data mining: concepts and techniques / Jiawei Han, Micheline Kamber, Jian Pei. – 3rd ed. ISBN 978-0-12-381479-1 “, OLAP is just a section of the (Data mining) book. OLAP summarizes/aggregates by the group while other data mining methods will try to find hidden patterns from non-aggregated data (= the detail is king). OLAP would approach managerial information as the summary is welcome, other data mining methods would approach research as the detail is required.

Great question @Evgeny Belenky. Thank you @Xavier Suriol ​for the answer.  OLAP consists of three basic analytical operations: consolidation (roll-up), drill-down, and slicing and dicing in the online mode, whereas “Data Mining” is a broader topic to mine massive data to find meaningful patterns out of the data by using various analytical techniques or framework such as CRISP-DM, Statical Techniques, Predictive Modeling techniques, exploratory data analysis, etc.  Data Mining can be done both in the online and offline mode. Sometimes, in data mining, there would not be any predefined objective. 

I have found both Knime and RapidMiner easy to use and powerful. That applies to producing numerical output, graphical representations, and PMML files. The latter helps if one wishes to produce diagnostic or recommender systems automatically, with no human intervention needed. I would, however, always have human oversight available and do a performance check occasionally. Both offer ways of integrating with Python or R should you wish to do something unusual. They are quick-to-learn for beginners. I often receive a dataset and do a quick analysis in, say, Knime before turning to SPSS for the heavier lifting.

So, most of our clients tend to like Qlik or Tableau for BI. It also depends on your data sources - if you're pulling from a newer cloud ERP there's generally more to work with in an easier fashion - but people also work wonders going straight to the data tables.  Power BI from Microsoft is a good option - mentioning regression analysis makes me think of Excel, and Power BI uses Excel as the user interface (sort of).  But depending on the ERP, you'll also want to take a gander at Host Analytics, Adaptive, Hyperion - not totally in the BI space, but may provide what you're looking for.  And between all of those is going to be a featureset that specializes in the data you're wanting to work with in the format you can access it. And how that clicks for you is the best indicator of what will end up being easiest for you to use.  Our practice for software selection is based on giving clients options between the best solutions, but it's so individual on the client side - we really don't know which software application will ring true in approach and featureset to the client.  We tend to make sure whatever choice they end up with is a proven product supported by solid, proven partners - because bad implementations of good software = junk.