Tag: model

Predictive Models Ain’t Easy to Deploy

 

This is a guest blog post by Carole Ann Matignon of Sparkling Logic. You can see more on Sparkling Logic at http://my.sparklinglogic.com/

Decision Management is about combining predictive models and business rules to automate decisions for your business. Insurance underwriting, loan origination or workout, claims processing are all very good use cases for that discipline… But there is a hiccup… It ain’t as easy you would expect…

What’s easy?

If you have a neat model, then most tools would allow you to export it as a PMML model – PMML stands for Predictive Model Markup Language and is a standard XML representation for predictive model formulas. Many model development tools let you export it without much effort. Many BRMS – Business rules Management Systems – let you import it. Tada… The model is ready for deployment.

What’s hard?

The problem that we keep seeing over and over in the industry is the issue around variables.

Those neat predictive models are formulas based on variables that may or may not exist as is in your object model. When the variable is itself a formula based on the object model, like the min, max or sum of Dollar amount spent in Groceries in the past 3 months, and the object model comes with transaction details, such that you can compute it by iterating through those transactions, then the problem is not “that” big. PMML 4 introduced some support for those variables.

The issue that is not easy to fix, and yet quite frequent, is when the model development data model does not resemble the operational one. Your Data Warehouse very likely flattened the object model, and pre-computed some aggregations that make the mapping very hard to restore.

It is clearly not an impossible project as many organizations do that today. It comes with a significant overhead though that forces modelers to involve IT resources to extract the right data for the model to be operationalized. It is a heavy process that is well justified for heavy-duty models that were developed over a period of time, with a significant ROI.

This is a show-stopper though for other initiatives which do not have the same ROI, or would require too frequent model refresh to be viable. Here, I refer to “real” model refresh that involves a model reengineering, not just a re-weighting of the same variables.

For those initiatives where time is of the essence, the challenge will be to bring closer those two worlds, the modelers and the business rules experts, in order to streamline the development AND deployment of analytics beyond the model formula. The great opportunity I see is the potential for a better and coordinated tuning of the cut-off rules in the context of the model refinement. In other words: the opportunity to refine the strategy as a whole. Very ambitious? I don’t think so.

About Carole Ann Matignon

http://my.sparklinglogic.com/index.php/company/management-team

Carole-Ann Matignon Print E-mail

Carole-Ann MatignonCarole-Ann Matignon – Co-Founder, President & Chief Executive Officer

She is a renowned guru in the Decision Management space. She created the vision for Decision Management that is widely adopted now in the industry.  Her claim to fame is managing the strategy and direction of Blaze Advisor, the leading BRMS product, while she also managed all the Decision Management tools at FICO (business rules, predictive analytics and optimization). She has a vision for Decision Management both as a technology and a discipline that can revolutionize the way corporations do business, and will never get tired of painting that vision for her audience.  She speaks often at Industry conferences and has conducted university classes in France and Washington DC.

She started her career building advanced systems using all kinds of technologies — expert systems, rules, optimization, dashboarding and cubes, web search, and beta version of database replication. At Cleversys (acquired by Kurt Salmon & Associates), she also conducted strategic consulting gigs around change management.

While playing with advanced software components, she found a passion for technology and joined ILOG (acquired by IBM). She developed a growing interest in Optimization as well as Business Rules. At ILOG, she coined the term BRMS while brainstorming with her Sales counterpart. She led the Presales organization for Telecom in the Americas up until 2000 when she joined Blaze Software (acquired by Brokat Technologies, HNC Software and finally FICO).

Her 360-degree experience allowed her to gain appreciation for all aspects of a software company, giving her a unique perspective on the business. Her technical background kept her very much in touch with technology as she advanced.

Oracle launches its version of R #rstats

From-

http://www.oracle.com/us/corporate/press/1515738

Integrates R Statistical Programming Language into Oracle Database 11g

News Facts

Oracle today announced the availability of Oracle Advanced Analytics, a new option for Oracle Database 11g that bundles Oracle R Enterprise together with Oracle Data Mining.
Oracle R Enterprise delivers enterprise class performance for users of the R statistical programming language, increasing the scale of data that can be analyzed by orders of magnitude using Oracle Database 11g.
R has attracted over two million users since its introduction in 1995, and Oracle R Enterprise dramatically advances capability for R users. Their existing R development skills, tools, and scripts can now also run transparently, and scale against data stored in Oracle Database 11g.
Customer testing of Oracle R Enterprise for Big Data analytics on Oracle Exadata has shown up to 100x increase in performance in comparison to their current environment.
Oracle Data Mining, now part of Oracle Advanced Analytics, helps enable customers to easily build and deploy predictive analytic applications that help deliver new insights into business performance.
Oracle Advanced Analytics, in conjunction with Oracle Big Data Appliance, Oracle Exadata Database Machine and Oracle Exalytics In-Memory Machine, delivers the industry’s most integrated and comprehensive platform for Big Data analytics.

Comprehensive In-Database Platform for Advanced Analytics

Oracle Advanced Analytics brings analytic algorithms to data stored in Oracle Database 11g and Oracle Exadata as opposed to the traditional approach of extracting data to laptops or specialized servers.
With Oracle Advanced Analytics, customers have a comprehensive platform for real-time analytic applications that deliver insight into key business subjects such as churn prediction, product recommendations, and fraud alerting.
By providing direct and controlled access to data stored in Oracle Database 11g, customers can accelerate data analyst productivity while maintaining data security throughout the enterprise.
Powered by decades of Oracle Database innovation, Oracle R Enterprise helps enable analysts to run a variety of sophisticated numerical techniques on billion row data sets in a matter of seconds making iterative, speed of thought, and high-quality numerical analysis on Big Data practical.
Oracle R Enterprise drastically reduces the time to deploy models by eliminating the need to translate the models to other languages before they can be deployed in production.
Oracle R Enterprise integrates the extensive set of Oracle Database data mining algorithms, analytics, and access to Oracle OLAP cubes into the R language for transparent use by R users.
Oracle Data Mining provides an extensive set of in-database data mining algorithms that solve a wide range of business problems. These predictive models can be deployed in Oracle Database 11g and use Oracle Exadata Smart Scan to rapidly score huge volumes of data.
The tight integration between R, Oracle Database 11g, and Hadoop enables R users to write one R script that can run in three different environments: a laptop running open source R, Hadoop running with Oracle Big Data Connectors, and Oracle Database 11g.
Oracle provides single vendor support for the entire Big Data platform spanning the hardware stack, operating system, open source R, Oracle R Enterprise and Oracle Database 11g.
To enable easy enterprise-wide Big Data analysis, results from Oracle Advanced Analytics can be viewed from Oracle Business Intelligence Foundation Suite and Oracle Exalytics In-Memory Machine.

Supporting Quotes

“Oracle is committed to meeting the challenges of Big Data analytics. By building upon the analytical depth of Oracle SQL, Oracle Data Mining and the R environment, Oracle is delivering a scalable and secure Big Data platform to help our customers solve the toughest analytics problems,” said Andrew Mendelsohn, senior vice president, Oracle Server Technologies.
“We work with leading edge customers who rely on us to deliver better BI from their Oracle Databases. The new Oracle R Enterprise functionality allows us to perform deep analytics on Big Data stored in Oracle Databases. By leveraging R and its library of open source contributed CRAN packages combined with the power and scalability of Oracle Database 11g, we can now do that,” said Mark Rittman, co-founder, Rittman Mead.
Oracle Advanced Analytics — an option to Oracle Database 11g Enterprise Edition – extends the database into a comprehensive advanced analytics platform through two major components: Oracle R Enterprise and Oracle Data Mining. With Oracle Advanced Analytics, customers have a comprehensive platform for real-time analytic applications that deliver insight into key business subjects such as churn prediction, product recommendations, and fraud alerting.

Oracle R Enterprise tightly integrates the open source R programming language with the database to further extend the database with Rs library of statistical functionality, and pushes down computations to the database. Oracle R Enterprise dramatically advances the capability for R users, and allows them to use their existing R development skills and tools, and scripts can now also run transparently and scale against data stored in Oracle Database 11g.

Oracle Data Mining provides powerful data mining algorithms that run as native SQL functions for in-database model building and model deployment. It can be accessed through the SQL Developer extension Oracle Data Miner to build, evaluate, share and deploy predictive analytics methodologies. At the same time the high-performance Oracle-specific data mining algorithms are accessible from R.

BENEFITS
  • Scalability—Allows customers to easily scale analytics as data volume increases by bringing the algorithms to where the data resides – in the database
  • Performance—With analytical operations performed in the database, R users can take advantage of the extreme performance of Oracle Exadata
  • Security—Provides data analysts with direct but controlled access to data in Oracle Database 11g, accelerating data analyst productivity while maintaining data security
  • Save Time and Money—Lowers overall TCO for data analysis by eliminating data movement and shortening the time it takes to transform “raw data” into “actionable information”
Oracle R Hadoop Connector Gives R users high performance native access to Hadoop Distributed File System (HDFS) and MapReduce programming framework.
This is a  R package
From the datasheet at

PMML Augustus

Here is a new-old system in open source for

for building and scoring statistical models designed to work with data sets that are too large to fit into memory.

http://code.google.com/p/augustus/

Augustus is an open source software toolkit for building and scoring statistical models. It is written in Python and its
most distinctive features are:
• Ability to be used on sets of big data; these are data sets that exceed either memory capacity or disk capacity, so
that existing solutions like R or SAS cannot be used. Augustus is also perfectly capable of handling problems
that can fit on one computer.
• PMML compliance and the ability to both:
– produce models with PMML-compliant formats (saved with extension .pmml).
– consume models from files with the PMML format.
Augustus has been tested and deployed on serveral operating systems. It is intended for developers who work in the
financial or insurance industry, information technology, or in the science and research communities.
Usage
Augustus produces and consumes Baseline, Cluster, Tree, and Ruleset models. Currently, it uses an event-based
approach to building Tree, Cluster and Ruleset models that is non-standard.

New to PMML ?

Read on http://code.google.com/p/augustus/wiki/PMML

The Predictive Model Markup Language or PMML is a vendor driven XML markup language for specifying statistical and data mining models. In other words, it is an XML language so that Continue reading “PMML Augustus”

Some Ways Anonymous Could Disrupt the Internet if SOPA is passed

This is a piece of science fiction. I wrote while reading Isaac Assimov’s advice to writers in GOLD, while on a beach in Anjuna.

1) Identify senators, lobbyists, senior executives of companies advocating for SOPA. Go for selective targeting of these people than massive Denial of Service Attacks.

This could also include election fund raising websites in the United States.

2) Create hacking tools with simple interfaces to probe commonly known software errors, to enable wider audience including the Occupy Movement students to participate in hacking. thus making hacking more democratic. What are the top 25 errors as per  http://cwe.mitre.org/cwss/

–http://www.decisionstats.com/top-25-most-dangerous-software-errors/ ?

 

Easy interface tools to check vulnerabilities would be the next generation to flooding tools like HOIC, LOIC – Massive DDOS atttacks make good press coverage but not so good technically

3) Disrupt digital payment mechanisms for selected targets (in step1) using tools developed in Step 2, and introduce random noise errors in payment transfers.

4) Help create a better secure internet by embedding Tor within Chromium with all tools for anonymity embedded for easy usage – a more secure peer to peer browser (like a mashup of Opera , tor and chromium).

or maybe embed bit torrents within a browser.

5) Disrupt media companies and cloud computing based companies like iTunes, Spotify or Google Music, just like virus, ant i viruses disrupted the desktop model of computing. After that offer solutions to the problems like companies of anti virus software did for decades.

6) Hacking websites is fine fun, but hacking internet databases and massively parallel data scrapers can help disrupt some of the status quo.

This applies to databases that offer data for sale, like credit bureaus etc. Making this kind of data public will eliminate data middlemen.

7) Use cross border, cross country regulatory arbitrage for better risk control of hacker attacks.

8) recruiting among universities using easy to use hacking tools to expand the pool of dedicated hacker armies.

9) using operations like those targeting child pornography to increase political acceptability of the hacker sub culture. Refrain from overtly negative and unimaginative bad Press Relations

10) If you cant convince  them to pass SOPA, confuse them 😉 Use bots for random clicks on ads to confuse internet commerce.

 

Topic Models

Some stuff on Topic Models-

http://en.wikipedia.org/wiki/Topic_model

In machine learning and natural language processing, a topic model is a type of statistical model for discovering the abstract “topics” that occur in a collection of documents. An early topic model was probabilistic latent semantic indexing (PLSI), created by Thomas Hofmann in 1999.[1] Latent Dirichlet allocation (LDA), perhaps the most common topic model currently in use, is a generalization of PLSI developed by David Blei, Andrew Ng, and Michael Jordan in 2002, allowing documents to have a mixture of topics.[2] Other topic models are generally extensions on LDA, such as Pachinko allocation, which improves on LDA by modeling correlations between topics in addition to the word correlations which constitute topics. Although topic models were first described and implemented in the context of natural language processing, they have applications in other fields such as bioinformatics.

http://en.wikipedia.org/wiki/Latent_Dirichlet_allocation

In statistics, latent Dirichlet allocation (LDA) is a generative model that allows sets of observations to be explained by unobserved groups that explain why some parts of the data are similar. For example, if observations are words collected into documents, it posits that each document is a mixture of a small number of topics and that each word’s creation is attributable to one of the document’s topics. LDA is an example of a topic model

David M Blei’s page on Topic Models-

http://www.cs.princeton.edu/~blei/topicmodeling.html

The topic models mailing list is a good forum for discussing topic modeling.

In R,

Some resources I compiled on Slideshare based on the above- Continue reading “Topic Models”

Business Analytics Projects

As per me, Analytics Projects get into these four  broad phases-

  • Business Problem  Phase– What needs to be done?
  1. Increase Revenues
  2. Cut Costs
  3. Investigate Unusual Events
  4. Project Timelines
  • Technical Problem Phase – Technical Problems in Project Execution 
  1. Data Availability /Data Quality/Data Augmentation Costs
  2. Statistical -(Technique based approach) , Hypothesis Formulation,Sampling, Iterations
  3. Programming-(Tool based approach) Analytics Platform Coding (Input, Formats,Processing)
  • Technical Solution Phase – Problem Solving using the Tools and Skills Available 
  1. Data Cleaning /Outlier Treatment/Missing Value Imputation
  2. Statistical -(Technique based approach) Error Minimization, Model Validation, Confidence Levels
  3. Programming-(Tool based approach) Analytics Platform Coding (Output, Display,Graphs)
  • Business Solution Phase– Put it all together in a word document, presentation and/or spreadsheet
  1. Finalized- Forecasts  , Models and Data Strategies
  2. Improvements  in existing processes
  3.  Control and Monitoring of Analytical Results post Implementation
  4. Legal and Compliance  guidelines to execution
  5. (Internal or External) Client Satisfaction and Expectation Management
  6. Audience Feedback based on presenting final deliverable to broader audience

Does the Internet need its own version of credit bureaus

Data Miners love data. The more data they have the better model they can build. Consumers do not love data so much and find sharing data generally a cumbersome task. They need to be incentivize for filling out survey forms , and for signing to loyalty programs. Lawyers, and privacy advocates love to use examples of improper data collection and usage as the harbinger of an ominous scenario. George Orwell’s 1984 never “mentioned” anything about Big Brother trying to sell you one more loan, credit card or product.

Data generated by customers is now growing without their needing to fill out forms and surveys. This data is about their preferences , tastes and choices and is growing in size and depth because it is generated from social media channels on the Internet.It is this data that can be and is captured by social media analytics.

Mobile data is also growing, including usage of location based applications and usage of Internet from the mobile phone is leading to further increases in data about consumers.Increasingly , location based applications help to provide a much more relevant context to the data generated. Just mobile data is expected to grow to 15 exabytes by 2015.

People want to have more and more conversations online publicly , share pictures , activity and interact with a large number of people whom  they have never met. But resent that information being used or abused without their knowledge.

Also the Internet is increasingly being consolidated into a few players like Microsoft, Amazon, Google  and Facebook, who are unable to agree on agreements to share that data between themselves. Interestingly you can use Yahoo as a data middleman between Google and Facebook.

At the same time, more and more purchases are being done online by customers and Internet advertising has grown much above the rate of growth of other mediums of communication.
Internet retail sales have the advantage that better demand predictability can lead to lower inventories as retailers need not stock up displays to look good. An Amazon warehouse need not keep material to simply stock up it shelves like a K-Mart does.

Our Hypothesis – An Analogy with how Financial Data Marketing is managed offline

  1. Financial information regarding spending and saving is much more sensitive yet the presence of credit bureaus alleviates these concerns.
  2. Credit bureaus collect information from all sources, aggregate and anonymize the individual components accordingly.They use SSN as a unique identifier.
  3. The Internet has a unique number too , called the Internet Protocol Address (I.P) 
  4. Should there be a unique identifier like Internet Security Number for the Internet to ensure adequate balance between the need for privacy as well as the need for appropriate targeting? 

After all, no one complains about privacy intrusions if their credit bureau data is aggregated , rolled up, and anonymized and turned into a propensity model for sending them direct mailers.

Advertising using Social Media and Internet

https://www.facebook.com/about/ads/#stories

1. A business creates an ad
Let’s say a gym opens in your neighborhood. The owner creates an ad to get people to come in for a free workout.
2. Facebook gets paid to deliver the ad
The owner sends the ad to Facebook and describes who should see it: people who live nearby and like running.
The right people see the ad
3. Facebook only shows you the ad if you live in town and like to run. That’s how advertisers reach you without knowing who you are.

Adding in credit bureau data and legislative regulation for anonymizing  and handling privacy data can expand the internet selling market, which is much more efficient from a supply chain perspective than the offline display and shop models.

Privacy Regulations on Marketing using Internet data
Should laws on opt out and do not mail, do not call, lists be extended to do not show ads , do not collect information on social media. In the offline world, you can choose to be part of direct marketing or opt out of direct marketing by enrolling yourself in various do not solicit lists. On the internet the only option from advertisements is to use the Adblock plugin if you are Google Chrome or Firefox browser user. Even Facebook gives you many more ads than you need to see.

One reason for so many ads on the Internet is lack of central anonymize data repositories for giving high quality data to these marketing companies.Software that can be used for social media analytics is already available off the shelf.

The growth of the Internet has helped carved out a big industry for Internet web analytics so it is a matter of time before social media analytics becomes a multi billion dollar business as well. What new developments would be unleashed in this brave new world is just a matter of time, and of course of the social media data!