Data Warehousing meant to serve Predictive Analytics?
Data Warehousing is a means to an end which is Data Mining (Both Descriptive & Predictive Analytics). The goal of Data Warehousing is only to provide a platform which is friendly for Data Mining. Yet most organisations restrict their Business Intelligence output to OLAP reporting only (Limited Descriptive Analytics). And with such limited vision, datawarehousese are designed keeping only OLAP in mind which doesnot make them friendlier for the bulk of Data Mining potential(Predictive Analysis) at times. Do you believe that Datawarehoues are misued and under utilized in this context?
Good Answers (3)
Gerald I
Principal Consulting Engineer at GE
Best Answers in: Equity Markets (3), Public Relations (2), Manufacturing (2), Computers and Software (2), Software Development (2), Wireless (2), Using LinkedIn (2), Venture Capital and Private Equity (1), Internationalization and Localization (1), Offshoring and Outsourcing (1), Criminal Law (1), Property Law (1), Advertising (1), Internet Marketing (1), Business Development (1), Business Analytics (1), Derivatives Markets (1), Hedge Funds (1), Non-profit Management (1), Quality Management and Standards (1), Personal Debt Management (1), Market Research and Definition (1), Professional Organizations (1), Ethics (1), Starting Up (1), Databases (1)
Atif,
Many corporations and their executives have not yet grasp the importance or even the power that can be obtained from Data Warehousing Analytics. From what I have read, this lack of understanding comes the not understanding the ROI involved with Data Warehousing. Since the return on investment is not immediately understood or comprehended, most executives will place very little resources and time in this IT model. This may explain the "Limited Descriptive Analytics" that your described in your dissertation and question.
There are several industries that are taking a more active role in learning and understanding the power of Data Warehousing. The links below point to some of these industries.
I hope this information helps you.
Links:
Patrick D
Computer & Security consultant, owner ProMind
Best Answers in: Computers and Software (6), Telecommunications (4), Computer Networking (2), Information Storage (2), Small Business (1), Enterprise Software (1), Wireless (1)
I agree fully with both previous answers.
The basis of the problem is the difference between the information needed for operational purposes (LDA) and for PA. In many cases systems (and therefore datastructures) are designed to produce operational output.
What is often overseen is that for datamining DA purposes additional info needs to be stored, in order to be able to categorize data in statistical usefull groups.
Since the people that provide the operational input and the people able to supply the addtional data are usully in different departments it is usually very hard to get both forms of information into a coherent dataset.
In many cases companies will even prefer to keep both completely separate and therefore create double input effort.
Another aspect is that it is often for performance reasons not indicated to run datamining applications on live production systems, so the CIO will no be eager to promote the concept. Especially since the specifications for the results of the DA will usally be vague, for the above mentioned reasons, and it will take quite a bit of time before enough data is stored to be useful for DA. This makes design and prototyping very precarious.
regards
Crystal N W
Intelligence Analyst at NSW Government
Best Answers in: Business Analytics (2), Facilities Management (1), Occupational Training (1), Risk Management (1), Planning (1), Career Management (1), Enterprise Software (1), Computers and Software (1)
One issue I'd like to raise is data security. From a technical point of view, the more gateways you have into your data warehouse the more vulnerable it is. So it's about finding a balance between letting the authorised users in and keeping the unauthorised users out. There's often a lot of very sensitive information stored so I can understand the restriction.
As an end user, there are always limitations I have to overcome to find a way to perform the best analysis, including restrictions on access and download. But I have to play devils advocate and say "a bad workman quarrels with his tools" - if you have data you need to perform PA and don't have it, or the interface isn't friendly enough to allow you to include it in your reports, then get proactive and suggest how giving it to you will benefit the organisation. But you should be willing to hear the other factors that might prevent it going ahead, that you may not be aware of at this point in time. The end user perspective is often the one with the least understanding of the variables in play in delivering to them.
More Answers (2)
Shahid H
Manager Data Warehouse at Telenor Pakistan
Best Answers in: Databases (2), Telecommunications (1)
Hi Atif,
As it is said that for an imaginative person the sky is limit, Same applies to Data Warehouse and its application for business. Looking at the five stages of a Data Warehouse (Reporting - > Analyzing -> Predicting->Operationalizing->Active Warehousing ), it clearly layes out the path for Data Warehouse growth. Its all about the visionary executives of a corporation who sets the direction and gradually moves teh data warehousing through these five stages.
Regards,
Shahid.
Atif, you seemed to have read my mind. I have noticed the same, that in some organizations datawarehouses are indeed misused. As Shahid said, Its all about the vision set by the executives of a corporation that dictates how the datawarehouse will be used. In my experience, most of those executives tend not to understand the detailed business requirments of the end user and therefore make these mistakes.
