Predictive Analysis with SQL Server 2008

This is a good article. In the interest of full-disclosure, I am familiar with Donald Farmer, Jamie MacLennan, Kamal Hathi and some of the individuals of the SQL data mining team within Microsoft. Also, I believe in the underlaying strategic principle that making predictive analytics available to all users within an enterprise is the keystone of turning data into actionable information for better decision-making. I know that the underlying predictive algorithms are based in good science and technology (as well as the products of SAS, SPSS, Business Objects, and many other vendors).

Two issues that I would like to explore with this product:
1. The interactive visualization of Analysis Services with Excel 2007. I believe that this is the key for data mining for the masses; and
2. The creation of industry-specific modules that allows for an easier designing of a multidimensional vector that is industry specific.

Family Study Associates Pesticide Use With Parkinson's Risk

This article shows the type of pattern that made me go into data mining. My father was a pure scientist (chemistry and mathematics) that dealt with pesticides in his research. Th entire laboratory staff (scientists and lab workers) all died of Parkinson's disease. The only thing that these people had in common was working with pesticides. These were very careful people that took extraordinary precautions in dealings with pesticides because they were aware of the harmful effects (even at the cellular level over a long period of time) of pesticides. My goal is to create a data mining visualization and predictive tool that can be used for as many people as possible to detect this type of patterns. Hopefully, when you have many people (experts as well as common people) looking at data patterns we are able to solve some of the serious problems that we are faced as a society.

TowerGroup: Losses Resulting from Mortgage Fraud in U.S. Will Reach $2.5 Billion in 2008

For U.S. companies the mortgage fraud software can be provided as a software as a service with analysts offshore. This will bring a cost-efficient and robust business analytics solution to this problem for financial institutions. As the Federal Reserve Board discuss the new oversight requirements for investment banks that are borrowing money from the Fed the requirement of a robust fraud prevention program could become a requirement.

Data Mining for Disease Management:Adding Value to Patient Records

This is a good article. I still question the underlying premise within the medical community that expects "perfect data" to equal perfect results. Data mining is about finding patterns (known and unknown) in the data that would allow imperfect people to make better decisions.

Software as a Service overview

This is a good overview video of software as a service (SaaS) if you need to explain the concept to others. In the case of SaaS analytics the issue is to create modules by industry. For example, I have worked on modules for payers, providers, pharmaceutical and DME companies. I am currently working on modules for the financial industry: banks, investment banks, brokerage companies, and insurance companies. Also, I am working modules for supply chain and biotechnology. Is anyone else working on SaaS modules by industry?

The Financial Market Crisis and Risks for Latin America

The purpose of this article is to give an example of the variables used in international credit risk exposure due to the changing world financial markets. We know that financial institutions in the U.S. and Europe have been negatively impacted, but banks in China and Latin America have a lower exposure because their main growth in organic or in their own domestic markets.

Therapeutic Cloning Works in Mice With Parkinson's

This is the type of discovery that a data mining system can have a direct impact.

Why good companies go bad

This is a classic article that I thought in this changing economic times could be a good reminder to companies that change could mean doing something different instead of just doing more or faster the same things that a company has done in the past. An enterprise analytics solution could help executives determine whether doing more of the same is helping or getting their companies in a bigger rut.

Enterprise Analytics: A Business Decision

The last three years have seen advances in efficient data mining algorithms, and computing advances that allow software companies to provide powerful analytical tools that were not available to the business community some years ago. Whether we refer to this type of software as data mining, predictive analytics, business intelligence, or analytics (web or business) their purpose is to efficiently detect patterns in large datasets that can lead to increase revenues or lower costs. The purpose of this article is to give a general framework to businesses regarding how to measure the different analytics solutions available in the marketplace.

First, let us make the difference between an analytical tool and an analytical solution. Most companies need an analytical solution instead of an analytical tool. Software vendors are in the business of selling analytical tools (SAS, Microsoft, Oracle, Business Objects, and SPSS). All these companies are using state-of-the-art science and technology to create powerful analytical tools. A hammer, a saw, and nails in the hands of a master carpenter can create a beautiful house. Those same tools in somebody else hands are something that you put in the garage. Therefore, the question is do you need an analytic tool or do you need an analytic solution. If your IT staff has not built an advanced analytical decision support system in the past what you need is a solution. If they have this experience then what you might need is a new tool. One way to tell the difference is whether the sum, or average (or mean) are the most common measurements used in your organization. If this is the case, you need an analytics solution and not a new tool.

Most carpenters will tell you that some people have all the tools needed to build a house in their garage. Therefore, before going out buying new tools you need to know: 1. what do you want to build? 2. What skills do you have in designing and building an analytical system? What tools you already have in your IT department?

Alignment with specific business objectives
What do you want to build? The answer to this question is that you want an analytical solution which is aligned with your organization strategic and operational objectives. This is one of the keystones of an analytical solution: it must match strategic and operational objectives. An enterprise strategic objective will represent the schematics or floor plan of your analytical solution. The enterprise operational objectives will represent the plumbing and the electrical systems of your analytical solution.

An analytical solution must ultimately represent the vision of executive management, while simultaneously be efficient to operate for those responsible for day-to-day operations. We know that a CEO can write his strategic vision even in a napkin and the role of his team should be to translate this vision into a series of strategic and operational goals. Hence, the importance that the overall strategic design of an analytics solution starts at the CEO and executive management team level. The decision to build an enterprise analytics system is an executive decision. A CEO thinks in terms of analytic solutions to business problems, and the IT department tendency is to think in terms of tools.

An experience analytics architect should be able to turn the executives input into an analytics solution prototype with some additional research and input from operations. This prototype should be detailed and must be approved by executive management before proceeding with specific business requirements and technical design. See,http://www.youtube.com/watch?v=bVmLfCajDjI.

Encompass all relevant data
Your data is the equivalent of your raw materials (wood, stone, bricks, pipes, flooring, carpets, and drywall). Since not all your materials are the analytics solutions architect and the lead software developers need to have the proper experience with large, diverse and complex datasets. Not all the houses are the same but the materials used are mostly the same. The same logic applies to the designing and building of an analytical solution. It is important not to confuse the quality of a tool to the quality of the materials. The tools can make a job easier for the builder, but is the quality of the materials in the hands of an expert builder that is going to make an average house an outstanding home.

The skill level of technical personnel varies and is as diverse as the number of home builders in the country. Some IT departments are made of technical staff that can give maintenance to current operations. Other IT departments have a specialized area for development, and others have fully functional project management offices (PMO). Very seldom do IT departments have an integrated analytics expertise (statistician and actuaries) and specialized software analytics developers (data cleansing, OLAP, and interactive visualization) within their departments that would allow the creation of advanced analytics solutions.

Flexibility of use for different business users
A robust analytical solution needs to be flexible for the different business users within an organization, and be adaptable to future analytical needs. Some users are interested in strategic objectives and others in operational objectives. An enterprise analytics solution should meet both strategic and operational demands. A keystone of an analytical solution is that it should allow different business users to have access to it. A business problem may have different potential solutions and analytical solution should take advantage of different perspectives from different users (human interaction) to identify potential solutions.

Also, it should have the capacity to accommodate future growth and changing business needs. The current business analytics requirements of an enterprise may be different from the analytical needs of the future. An analytics solution needs to be design with the capacity to growth, and the flexibility to accommodate unforeseen business issues.

In conclusion, a business analytics solution should cover the fundamentals:
1. Emphasize the alignment of the strategic and operational objectives to the analytic solution instead of the analytical tool;
2. Make sure that you have the correct materials (staff and data) to build your solution;and
3. Design a flexible analytical solution that could be used by strategic and operational users.

Human Computer Interaction

This is the site for the Human Computer Interaction at the University of Konstanz in Germany. If you want to know about best practices in interactions and visualization techniques, see http://hci.uni-konstanz.de/index.php?a=teaching&b=corner&c=15851838&lang=en

Interaction Techniques for High Resolution Displays

A video of interactive visualization. Imagine analyzing data using this type of techniques and that is available to everybody in an organization.

Real Time Scalable Visual Analysis on Mobile Devices

This is the next logical step: integrating analytics visualization with mobile devices.

Gartner: Emerging Technologies Will Help Drive Mainstream BI Adoption

Gartner's analysis is 100% on target. Business intelligence is more than creating an enterprise data warehouse, it is about transforming data into actionable information by allowing all individuals within an organization to make decisions based on business analytics.

Mid-Market Insurance Provider Selects Blink Logic for SaaS BI Solution

Two main reasons that I am posting this article:
1. It illustrates that analytics can be provided using a software as a service (SaaS) business model; and
2. It brings home the point that business analytics is not only for highly-skilled analysts, but available to the decision-makers as well.
This is the future of analytics!

Human Factors In Visualization Research

It is the interaction of the computing capabilities and the human mind capabilities that allows for evidence-based decision making in situations involving large sets of data. Therefore, in order to make evidence-based decisions data must be summarized in such a way that takes advantage of the increddible capabilities of the human mind. A good data mining visualization tool should be able to separate the clusters of data and clarify the driving factors by clearly and cognitively recognizing the patterns or trends in the data.

This is a good article that emphasizes the cooperation between the end-users and the development of visualization tools for data mining and business analytics. The bottom line: make it simple and intuitive!

Visualization Projects for Data Mining

These are the current projects for the visualization group at the Lawrence Berkeley National Laboratory. One of the members of the Business Analytics Group expressed interest in the latest developments in network traffic analysis and cybersecurity (see, http://vis.lbl.gov/Vignettes/QDV-NetworkTraffic/qdv-vignette.html).
I would suggest to spend 1-2 minutes for each project and just look at the visualizations and determine if this is something that can be apply to an organization as part of your business analytics and predictive modeling. The visualization technology is available right now. The question is whether the business is in a position to accept these concepts as a way to look at large datasets. We know that there is not even a probability that a spreadsheet can capture the complexity of large datasets. You may want to start with proposing a simpler visualization like SAS, SPSS and Cognos, Business Objects, or Microsofot Analysis Services and Excel 2007. The key for these visualization tools is to train the business users in using these tools. My approach is to teach first executive management and find what tools they find more useful. The purpose of these visualization tools is to let the human brain look at the totality of the data and start discovering new trends and patterns.

SAS Advances Enterprise Intelligence

Without a visualization example this are just words.

Risk Management Confidence: Higher-ups More Conservative

My interpretation of this article: executives that use predictive modeling understand the complexity of the risks involved and threfore are less prone to take non-fact based risks.

MADE IN IBM LABS: IBM Software Finds Hidden Product and Service Insight in Customer Interactions

I wish that when this type of news would come out a small visualization example of the output will be publish, since the visualization is what is going to tell us whether a business analytics tool is efficient in assisting an end user in looking at the totality of the data by separating the clusters in the data and clarifying the driving factors.

Financial Markets Crisis: A Potential Solution

In the late 1980's I had the opportunity to participate in the discussions of what to do with billions of dollars in commercial bank debt for countries in Latin America. In a nutshell, commercial banks had acquired a very large exposure to debt from Latin American countries, and in reality most of that debt was worthless although the interest rates were very high. The issue was similar than today's mortage backed subprime securities: the lending institution failed to follow fundamentally sound practices in their portfolios and their exposure came to the point of affecting the world finance markets. I remember the 2nd International Conference on Debt and Trade where the policy makers, financial instituions, and debt-ridden countries met. One of the solutions was to create markets for this exposure and hence the birth of the Bradley bonds. At the time, I was a proponent of a more drastic measure, to let the market forces decide the outcome of each financial instituion according to their exposure. Over the years I have come to the conclusion that the Bradley bond secondary market has substantially assisted in bringing order to the financial markets in a way that has not negatively affected Latin America countries.

My proposed solution to the current financial markets crisis is a three tier solution:
1. Allow a market for mortage backed securities in which central banks around the world could buy some of the exposures of large financial institutions. I would suggest up to 3% of the market capitalization of each large financial institution. This solution will allow the internalization of the financial markets by allowing central banks to have a direct ownership in private financial institutions. I know that this is a radical solution, but the liquidity problems in the financial markets are of such magnitutde that it requires the intervention of central banks around the world in a different way and at a different magnitude that past remedies. Special class of stocks can be created that will allow foreign central banks to participate that would restric voting rights, while simultaneously allows their input into investment practices.
2. Create a tax incentive for financial institutions in which they could lower by a total up to 3% the interest rate and/or the principal of the mortgage loan, for properties with the value of a mortage of up to $250,000. The total amount of the tax incentive could be cap at $2,500 per qualifying loan.
3. For properties with a mortgage value greater than $250,000 I would suggest to let the market forces dictate what will happen to these properties and the financial institutions that made those investments.

DATA MINING IN BANKING AND FINANCE: A NOTE FOR BANKERS

This article covers the fundamentals of data mining in financial markets and banking. The section on risk management (financial market risk and credit risk) is something that is worth taking a look at the fundamentals in today's changing financial markets. It is amazing how many times we need to go back to the fundamentals in the banking and finance industries to make sense of the corrections in the markets.

Business Analytics and Financial Markets Liquidity Issues

The news that Bear Sterns needs liquidity assistance from JP Morgan Chase and the Federal Reserve was predictable using off-the-shelve predictive analytics software. For those who are engaged in financial markets analytics let me suggest that instead of using Bear Sterns as your sampling data (before applying in to the entire dataset), use instead the Carlyle Fund data as your market for predictions. This data is less noisy and would help you determine the ratio of mortage based secutirites: total capital under management to determine risk in liquidity.

Also, please look at the other variables (i.e., geography, diversification, currency hedges, etc.) to determine what constitutes a robust liquidity scenario. In other words, invert the scenario and analyze the driving factors. Let me suggest the creation of multidimensional vectors and combine them with cluster analysis so you can separate the clusters of data and clarify the driving factors. This is a great opportunity to bring potential solutions to the executives.

A technical framework for sense-and-respond business management

This paper from the IBM Labs presents a good technical framework about what processes could be needed to achieve change managemetn when introducing predictive analytics (phrase taken from James Taylor) in a rapid changing enviroment. I am not suggesting that this process is exclusive or innovative, but I am suggesting that a process is indeed needed for change management to take place (if you fail to plan, you plan to fail). What other processes, or components of a process, are needed to introduce predictive analytics into a rapid changing environment? If you take into account the changing nature of the world economy and how is affecting organizations, what processes need to be in place to introduce predictive analytics into an organization? We know that is not a matter of the science and technology since those are already in place.

Emerging Trends in Business Analytics

I specifically like the emphasis and analysis of business users as it concerns business analytics in this article. The emphasis on solving business problems (not technical problems) and the ability to measure results are great points too. This brings me to the topic of change management and metrics. A collegue of mine, Mitch Weisberg, is the expert in this area: how to have the right metrics and the right processes that will incorporate advanced business analytics into the fabric of an organization.

Top 10 Unsolved Information Visualization Problems

A well thought out paper about the issues with visualization of information. I believe that the issue has been defined by a collegue, Edith Ohri, when she told me that that we (those with the expertise) have the responsibility to explain the issues to the executives in such a way that they can understand. Her definition of multidimensional vector analysis is beautiful: separating the clusters of data while clarifying the driving factors.

Cognos and SPSS Forge Partnership to Deliver Predictive Analytics Integration

This is big news. This moves Cognos from a reporting tool to a predictive analytics visualization tool. This bring IBM and SPSS in direct competition with Microsft, SAP, and SAS.

Information Visualization and Visual Data Mining

A little thick but extremely accurate. As companies prepare to provide analytics in software as a service (SaaS) they must be aware of the nteraction between data mining visualization and the human component. Data mining visualization is the next evolutionary step in utilizing the full capabilities of the human mind.

Business intelligence on demand

The next generation of predictive analytics is here!
Congratultions for a job well done!
Software as a Service (SaaS) in predictive analytics!

Why Don't American CIOs Want to Lead In Emerging Technologies?

Read this article. My answers to the question: timidity, lack of understanding of what it means to be an executive, some people are followers and some are leaders, and it is the safe route.

Why Business Analytics as a Service Won’t Spook IT

This is a very good and insightful article. Business Analytics as a Service (SaaS) is where the market is moving to. the main reason is that CIOs and business users do not want the details of advanced analytics, they just want a tool that works. Last week I was talking with a couple of IT executives from different large companies and they are looking for something "actionable". SaaS meets this requirement.

My only difference with the authors is that when people talk about "Excel hell" they do not know the capabilities that SQL 2005 Server Analysis Services and Excel 2007 can bring to an organization in terms of advanced analytics. I can see the combination of these two products becoming the first sucessful SaaS product for large enterprises as well as small businesses.

Visualization of Data Mining

I have found that using a clustering analysis visualization including a multidimensional vector is an efficient way to cut the cost of predictive modeling while simultaneously making data mining available to everyone within an organization. The advantage of this process is that it allows everyone to see large data sets in such a way that they can draw their own conclusions even without subject matter expertise.