Search Terms for Data Management & Analytics

Recently, for a prospective customer, I created a list of some search terms to provide them with some "late night" reading on data management & analytics. I've tried these terms out on Google, and as suspected, for most, the first hit is for Wikipedia. While most articles in Wikipedia need to be taken with a grain of salt, they will give you a good overview. [By the way, I use the "Talk" page on the articles to see the discussion and arguments about the article's content as an indicator of how big a grain of salt is needed for that article] &#59;) So plug these into your favorite search engine, and happy reading.

  • Reporting - top two hits on Google are Wikipedia, and, interestingly, Pentaho
  • Ad-hoc reporting
  • OLAP - one of the first page hits is for Julian Hyde's blog, creator of the open source tool for OLAP, Mondrian, as well as real-time analytics engine, SQLstream
  • Enterprise dashboard - interestingly, Wikipedia doesn't come up in the top hits for this term on Google, so here's a link for Wikipedia: http://en.wikipedia.org/wiki/Dashboards_(management_information_systems)
  • Analytics - isn't very useful as a search term, but the product page from SAS gives a nice overview
  • Advanced Analytics - is mostly marketing buzz, so be wary of anything that you find using this as search term

Often, Data Mining, Machine Learning and Predictives are used interchangeably. This isn't really correct, as you can see from the following five search terms…

  • Data Mining
  • Machine Learning
  • Predictive Analytics
  • Predictive Intelligence - is an earlier term for Predictives that has mostly been supplanted by Predictive Analytics. I actually prefer just "Predictives".
  • PMML - Predictive Modeling Markup Language - is a way of transporting predictive models from one software package to another. Few packages will both export and import PMML. The lack of that capability can lock you into a solution, making it expensive to change vendors. The first hit for PMML on Google today is the Data Mining Group, which is a great resource. One company listed, Zementis, is a start-up that is becoming a leader in running data mining and predictive models that have been created anywhere
  • R - the R statistical language, is difficult to search on Google. Go to http://www.r-project.org/ and http://www.rseek.org/ … instead. R is useful for writing applications for any type of statistical analysis, and is invaluable for creating new algorithms and predictive models
  • ETL - Extract, Transform & Load, is the most common way of getting information from source systems to analytic systems
  • ReSTful Web Services - Representational State Transfer - can expose data as a web service using the four verbs of the web
  • SOA
  • ADBMS - Analytic Database Management Systems doesn't work well as a search term. Start with the Eigenbase.org site and follow the links from the Eigenbase subproject, LucidDB. Also, check out AsterData
  • Bayes - The Reverend Thomas Bayes came up with this interesting approach to statistical analysis in the 1700s. I first started creating Bayesian statistical methods and algorithms for predicting reliability and risk associated with solid propellant rockets. You'll find good articles using Bayes as a search term in Google. A bit denser article can be found at http://www.scholarpedia.org/article/Bayesian_statistics And some interesting research using Bayes can be found at: Andrew Gelman's Blog. You're likely familiar with one common Bayesian algorithm, naïve Bayes, which is used by most anti-spam email programs. Other forms are objective Bayes with non-informative priors and the original Subjective Bayes. I have an old aerospace joke about the Rand Corporation's Delphi method, based on subjective Bayes :-) I created my own methodology, and don't really care for naïve Bayes nor non-informative priors.
  • Sentiment Analysis - which is one of Seth Grimes' current areas of research
  • Decision Support Systems - in addition to searching on Google, you might find my recent OSS DSS Study Guide of interest

Let me know if I missed your favorite search term for data management & analytics.

Data Artisan Smith or Scientist

Over the past few months, a debate has been proceeding on whether or not a new discipline, a new career path, is emerging from the tsunami of data bearing down on us. The need for a new type of Renaissance [Wo]Man to deal with the Big Data onslaught. To whit, Data Science.

I'm writing about this now, because last night, at an every-three-week get together devoted to cask beer and data analysis, the topic came up. [Yes, every-THREE-weeks - a month is too long to go without cask beer fueled discussions of Rstats, BigData, Streaming SQL, BI and more.] The statisticians in the group, including myself, strongly disagreed with the way the term is being used; the software/database types were either in favor or ambivalent. We all agreed that a new, interdisciplinary approach to Big Data is needed. Oh, and I'll stay on topic here, and not get into another debate as to the definition of "Big Data". &#59;)

This lively conversation reinforced my desire to write about Data Science that swelled up in me after reading "What is Data Science?" by Mike Loukides published on O'Reilly Radar, and a subsequent discussion on Twitter held the following weekend, concerning data analytics.

The term "Data Science" isn't new, but it is taking on new meanings. The Journal of Data Science published JDS volume 1, issue 1 in January of 2003. The Scope of the JDS is very clearly related to applied statistics

By "Data Science", we mean almost everything that has something to do with data: Collecting, analyzing, modeling...... yet the most important part is its applications --- all sorts of applications. This journal is devoted to applications of statistical methods at large.
-- About JDS, Scope, First Paragraph

There is also the CODATA Data Science Journal, which appears to have last been updated in August of 2007, and currently has no content, other than its self-description as

The Data Science Journal is a peer-reviewed electronic journal publishing papers on the management of data and databases in Science and Technology.

I think that two definitions can be derived from these two journals.

  1. Data Science is systematic study, through observation and experiment, of the collection, modeling, analysis, visualization, dissemination, and application of data.
  2. Data Science is the use of data and database technology within physical and natural sciences and engineering.

I can agree with the first, especially with the JDS Scope clearly stating that Data Science is applied statistics.

The New Oxford American Dictionary, on which the Apple Dictionary program is based, defines science as a noun

the intellectual and practical activity encompassing the systematic study of the structure and behaviour of the physical and natural world through observations and experiments.

And a similar definition of science can be found on Dictionary.com.

In many ways, I like Mike Loukides' article "What is Data Science?" in how it highlights the need for this new discipline. I just don't like what he describes to be the new definition of "data science". Indeed, I very much disagree with this statement from the article.

Using data effectively requires something different from traditional statistics, where actuaries in business suits perform arcane but fairly well-defined kinds of analysis. What differentiates data science from statistics is that data science is a holistic approach. We're increasingly finding data in the wild, and data scientists are involved with gathering data, massaging it into a tractable form, making it tell its story, and presenting that story to others.

A statistician is not an actuary. They're very different roles. I know this because I worked for over a decade applying statistics to determining the reliability and risk associated with very large, complex systems such as rockets and space-borne astrophysics observatories. I once hired a Cal student as an intern because she feared that the only career open to her as a math major, was to be an actuary. I showed her a different path. So, yes, I know, from experience, that a statistician is not an actuary. Actually, the definition of a data scientist given, that is "gathering data, massaging it into a tractable form, making it tell its story, and presenting that story to others" is exactly what a statistician does.

I do however see the need for a new discipline, separate from applied statistics, or data science. The massive amount of data to come from an instrumented world with strongly interconnected people and machines, and real-time analysis, inference and prediction from those data, will require inter-disciplinary skills. But I see those skills coming together in a person who is more of a smith, or, as Julian Hyde put it last night, an artisan. Falling back on the old dictionary again, a smith is someone who is skilled in creating something with a specific material; an artisan is someone who is skilled in a craft, making things by hand.

Another reason that I don't like the term "data science" for this interdisciplinary role, stems from what Mike Loukides describes in his article "What is Data Science?" as the definition for this new discipline "Data science requires skills ranging from traditional computer science to mathematics to art". I agree that the new discipline requires these three things, and more, even softer skills. I disagree that these add up to data science.

I even prefer "data geek", as defined by Michael E. Driscoll in "The Three Sexy Skills of Data Geeks". Michael Driscoll's post of 2009 May 27 certainly agrees skill-wise with Mike Loukides post of 2010 June 02.

  1. Skill #1: Statistics (Studying)
  2. Skill #2: Data Munging (Suffering)
  3. Skill #3: Visualization (Storytelling)

And I very much prefer "Data Munging" to "Computer Science" as one of the three skills.

I'll stick to the definition that I gave above for data science as "systematic study, through observation and experiment, of the collection, modeling, analysis, visualization, dissemination, and application of data". This is also applied statistics. So, what else is needed for this new discipline? Well, Mike and Michael are correct: computer skills, especially data munging, and art. Well, any statistician today has computer skills, generally in one or more of SAS, SPSS, R, S-plus, Python, SQL, Stata, MatLab and other software packages, as well as familiarity with various data storage & management methods. Some statisticians are even artists, perhaps as story tellers, as evidenced by that rare great teacher or convincing expert witness, perhaps as visualizers, creating statistically accurate animations to clearly describe the analysis, as evidenced by the career of that intern I hired so many years ago.

The data smith, the data artisan, must be comfortable with all forms of data:

  • structured,
  • unstructured and
  • semi-structured

Just as any other smith, someone following this new discipline might serve an apprenticeship creating new things from these forms of data such as a data warehouse or an OLAP cube, a sentiment analysis or a streaming SQL sensor web, or a recommendation engine or complex system predictives. The data smith must become very comfortable with putting all forms of data together in new ways, to come to new conclusions.

Just as a goldsmith will never make a piece of jewelry identical to the one finished days before, just as art can be forged but not duplicated, the data smith, the data artisan will glean new inferences every time they look at the data, will make new predictions with every new datum, and the story they tell, the picture they paint, will be different each time.

And perhaps then, the data smith becomes a master, an artisan.

PS: Here's a list of links to that Twitter conversation among some of the most respected people in the biz, on Data Analytics

  1. https://twitter.com/NeilRaden/status/15512935981
  2. https://twitter.com/NeilRaden/status/15513225191
  3. https://twitter.com/NeilRaden/status/15513275261
  4. https://twitter.com/NeilRaden/status/15513453916
  5. https://twitter.com/datachick/status/15513460384
  6. https://twitter.com/NeilRaden/status/15513488053
  7. https://twitter.com/datachick/status/15513677836
  8. https://twitter.com/CMastication/status/15513772446
  9. https://twitter.com/NeilRaden/status/15513821393
  10. https://twitter.com/NeilRaden/status/15513854916
  11. https://twitter.com/NeilRaden/status/15513915694
  12. https://twitter.com/alecsharp/status/15513980301
  13. https://twitter.com/NeilRaden/status/15514104372
  14. https://twitter.com/alecsharp/status/15514097194
  15. https://twitter.com/CMastication/status/15514374095
  16. https://twitter.com/estrenuo/status/15514634644
  17. https://twitter.com/NeilRaden/status/15515243453
  18. https://twitter.com/CMastication/status/15516185085
  19. https://twitter.com/annmariastat/status/15516321715
  20. https://twitter.com/NeilRaden/status/15519544709
  21. https://twitter.com/NeilRaden/status/15519597061
  22. https://twitter.com/NeilRaden/status/15519621974
  23. https://twitter.com/skemsley/status/15519932631
  24. https://twitter.com/skemsley/status/15519932631
  25. https://twitter.com/aristippus303/status/15520146540
  26. https://twitter.com/NeilRaden/status/15520478566
  27. https://twitter.com/SethGrimes/status/15520765766
  28. https://twitter.com/SethGrimes/status/15520851678
  29. https://twitter.com/NeilRaden/status/15521050387
  30. https://twitter.com/NeilRaden/status/15521106901
  31. https://twitter.com/NeilRaden/status/15521133647
  32. https://twitter.com/NeilRaden/status/15521192977
  33. https://twitter.com/SethGrimes/status/15521579977
  34. https://twitter.com/ryanprociuk/status/15521637974

Technology for the OSS DSS Study Guide

'Tis been longer than intended, but we finally have the technology, time and resources to continue with our Open Source Solutions Decision Support System Study Guide (OSS DSS SG).

First, I want to thank SQLstream for allowing us to use SQLstream as a part of our solution. As mentioned in our "First DSS Study Guide" post, we were hoping to add a real-time component to our DSS. SQLstream is not open source, and not readily available for download. It is however, a co-founder and core contributer to the open source Eigenbase Project, and has incorporated Eigenbase technology into its product. So, what is SQLstream? To quote their web site, "SQLstream enables executives to make strategic decisions based on current data, in flight, from multiple, diverse sources". And that is why we are so interested in having SQLstream as a part of our DSS technology stack: to have the capability to capture and manipulate data as it is being generated.

Today, there are two very important classes of technologies that should belong to any DSS: data warehousing (DW) and business intelligence (BI). What actually comprises these technologies is still a matter of debate. To me, they are quite interrelated and provide the following capabilities.

  • The means of getting data from one or more sources to one or more target storage & analysis systems. Regardless of the details for the source(s) and the target(s), the traditional means in data warehousing is Extract from the source(s), Transform for consistency & correctness, and Load into the target(s), that is, ETL. Other means, such as using data services within a services oriented architecture (SOA) either using provider-consumer contracts & Web Service Definition Language (WSDL) or representational state transfer (ReST) are also possible.
  • Active storage over the long term of historic and near-current data. Active storage as opposed to static storage, such as a tape archive. This storage should be optimized for reporting and analysis through both its logical and physical data models, and through the database architecture and technologies implemented. Today we're seeing an amazing surge of data storage and management innovation, with column-store relational database management systems (RDBMS), map-reduce (M-R), key-value stores (KVS) and more, especially hybrids of one or several of old and new technologies. The innovation is coming so thick and fast, that the terminology is even more confused than in the rest of the BI world. NoSQL has become a popular term for all non-RDBMS, and even some RDBMS like column-store. But even here, what once meant No Structured Query Language now is often defined as Not only Structured Query Language, as if SQL was the only way to create an RDBMS (can someone say Progress and its proprietary 4GL).
  • Tools for reporting including gathering the data, performing calculations, graphing, or perhaps more accurately, charting, formating and disseminating.
  • Online Analytical Processing (OLAP) also known as "slice and dice", generally allowing forms of multi-dimensional or pivot analysis. Simply put, there are three underlying concepts for OLAP: the cube (a.k.a. hypercube, multi-dimensional database [MDDB] or OLAP engine), the measures (facts) & dimensions, and aggregation. OLAP provides much more flexibility than reporting, though the two often work hand-in-hand, especially for ad-hoc reporting and analysis.
  • Data Mining, including machine learning and the ability to discover correlations among disparate data sets.

For our purposes, an important question is whether or not there are open source, or at least open source based, solutions for all of these capabilities. The answer is yes. As a matter of fact, there are three complete open source BI Suites [there were four, but the first, written in PERL, the Bee Project from the Czech Republic, is no longer being updated]. Here's a brief overview of SpagoBI, JasperSoft, and Pentaho.

Capability SpagoBI JasperSoft Pentaho
ETL Talend Talend
JasperETL
KETTLE
PDI
Included
DBMS
HSQLDB
MySQL
Reporting BIRT
JasperReport
JasperReports
iReports
jFreeReports
Analyzer jPivot
PaloPivot
JasperServer
JasperAnalysis
jPivot
PAT
OLAP Mondrian Mondrian Mondrian
Data Mining Weka None Weka

We'll be using Pentaho, but you can use any of the these, or any combination of the OSS projects that are used by these BI Suites, or pick and choose from the more than 60 projects in our OSS Linkblog, as shown in the sidebar to this blog. All of the OSS BI Suites have many more features than shown in the simple table above. For example, SpagoBI has good tools for geographic & location services. Also, JasperSoft Professional and Enterprise Editions have many features than their Community Edition, such as Ad Hoc Reporting and Dashboards. Pentaho has a different Analyzer in their Enterprise Edition than either jPivot or PAT, Pentaho Analyzer, based upon the SaaS ClearView from the now-defunct LucidEra, as well as ease-of-use tools such as an OLAP schæma designer, and enterprise class security and administration tools.

Data warehousing using general purpose RDBMS systems such as Oracle, EnterpriseDB, PostrgeSQL or MySQL, are gradually giving way to analytic database management system (ADBMS), or, as we mentioned above, the catch-all NoSQL data storage systems, or even hybrid systems. For example, Oracle recently introduced hybrid column-row store features, and Aster Data has a column-store Massive Parallel Processing (MPP) DBMS|map-reduce hybrid [updated 20100616 per comment from Seth Grimes]. Pentaho supports Hadoop, as well as traditional general purpose RDBMS and column-store ADMBS. In the open source world, there are two columnar storage engines for MySQL, Infobright and Calpont InfiniDB, as well as one column-store ADBMS purpose built for BI, LucidDB. We'll be using LucidDB, and just for fun, may throw some data into Hadoop.

In addition, a modern DSS needs two more primary capabilities. Predictives, sometimes called predictive intelligence or predictive analytics (PA), which is the ability to go beyond inference and trend analysis, assigning a probability, with associated confidence, or likelihood of an event occurring in the future, and full Statistical Analysis, which includes determining the probability density or distribution function that best describes the data. Of course, there are OSS projects for these as well, such as The R Project, the Apache Common Math libraries, and other GNU projects that can be found in our Linkblog.

For statistical analysis and predictives, we'll be using the open source R statistical language and the open standard predictive model markup language (PMML), both of which are also supported by Pentaho.

We have all of these OSS projects installed on a Red Hat Enterprise Linux machine. The trick will be to get them all working together. The magic will be in modeling and analyzing the data to support good decisions. There are several areas of decision making that we're considering as examples. One is fairly prosaic, one is very interesting and far-reaching, and the others are somewhat in between.

  1. A fairly simple example would be to take our blog statistics, a real-time stream using SQLstream's Twitter API, and run experiments to determine whether or not, and possibly how, Twitter affects traffic to and interaction with our blogs. Possibly, we could get to the point where we can predict how our use of Twitter will affect our blog.
  2. A much more far-reaching idea was presented by Ken Winnick to me, via Twitter, and has created an on-going Twitter conversation and hashtag, #BPgulfDB. Let's take crowd sourced, government, and other publicly available data about the recent oilspill in the Gulf of Mexico, and analyze it.
  3. Another idea is to take historical home utility usage plus current smart meter usage data, and create a real-time dashboard, and even predictives, for reducing and managing energy usage.
  4. We also have the opportunity of using public data to enhance reporting and analytics for small, rural and research hospitals.

OSS DSS Formalization

The next step in our open source solutions (OSS) for decision support systems (DSS) study guide (SG), according to the syllabus, is to make our first decision: a formal definition of "Decision Support System". Next, and soon, will be a post listing the technologies that will contribute to our studies.

The first stop in looking for a definition of anything today, is Wikipedia. And indeed, Wikipedia does have a nice article on DSS. One of the things that I find most informative about Wikipedia articles, is the "Talk" page for an article. The DSS discussion is rather mild though, no ongoing debate as can be found on some other talk pages, such as the discussion about Business Intelligence. The talk pages also change more often, and provide insight into the thoughts that go into the main article.

And of course, the second stop is a Google search for Decision Support System; a search on DSS is not nearly as fruitful for our purposes. :)

Once upon a time, we might have gone to a library and thumbed through the card catalog to find some books on Decision Support Systems. A more popular approach today would be to search Amazon for Decision Support books. There are several books in my library that you might find interesting for different reasons:

  1. Pentaho Solutions: Business Intelligence and Data Warehousing with Pentaho and MySQL by Roland Bouman & Jos van Dongen provides a very good overview of data warehousing, business intelligence and data mining, all key components to a DSS, and does so within the context of the open source Pentaho suite
  2. Smart Enough Systems: How to Deliver Competitive Advantage by Automating Hidden Decisions by James Taylor & Neil Raden introduces business concepts for truly managing information and using decision support systems, as well as being a primer on data warehousing and business intelligence, but goes beyond this by automating the data flow and decision making processes
  3. Business Intelligence Roadmap: The Complete Project Lifecycle for Decision-Support Applications by Larissa T. Moss & Shaku Atre takes a business, program and project management approach to implementing DSS within a company, introducing fundamental concepts in a clear, though simplistic level
  4. Competing on Analytics: The New Science of Winning by Thomas H. Davenport & Jeanne G. Harris in many ways goes into the next generation of decision support by showing how data, statistical and quantitative analysis within a context specific processes, gives businesses a strong lead over their competition, albeit, it does so at a very simplistic, formulaic level

These books range from being technology focused to being general business books, but they all provide insight into how various components of DSS fit into a business, and different approaches to implementing them. None of them actually provide a complete DSS, and only the first focuses on OSS. If you followed the Amazon search link given previously, you might also have noticed that there are books that show Excel as a DSS, and there is a preponderance of books that focus on the biomedical/pharmaceutical/healthcare industry. Another focus area is in using geographic information systems (actually one of the first uses for multi-dimensional databases) for decision support. There are several books in this search that look good, but haven't made it into my library as yet. I would love to hear your recommendations (perhaps in the comments).

From all of this, and our experiences in implementing various DW, BI and DSS programs, I'm going to give a definition of DSS. From a previous post in this DSS SG, we have the following:

A DSS is a set of processes and technology that help an individual to make a better decision than they could without the DSS.
-- Questions and Commonality

As we stated, this is vague and generic. Now that we've done some reading, let's see if we can do better.

A DSS assists an individual in reaching the best possible conclusion, resolution or course of action in stand-alone, iterative or interdependent situations, by using historical and current structured and unstructured data, collaboration with colleagues, and personal knowledge to predict the outcome or infer the consequences.

I like that definition, but your comments will help to refine it.

Note that we make no mention of specific processes, nor any technology whatsoever. It reflects my bias that decisions are made by individuals not groups (electoral systems not withstanding). To be true to our "TeleInterActive Lifestyle" &#59;) I should point out that the DSS must be available when and where the individual needs to make the decision.

Any comments?

R the next Big Thing or Not

Recently, AnnMaria De Mars, PhD (multiple) and Dr. Peter Flom, PhD have stirred up a bit of a tempest in a tweet-pot, as well as in the statistical blogosphere, with comparisons of R and SAS, IBM/SPSS and the like. I've commented on both of their blogs, but decided to expand a bit here, as the choice of R is something that we planned to cover in a later post to our Open Source Solutions Decision Support Systems Study Guide. First, let me say that Dr. De Mars and Dr. Flom appear to have posted completely independently of each other, and further, that their posts have different goals.

In The Next Big Thing, Dr. De Mars is looking for the next big thing, both to keep her own career on-track, and to guide students into areas of study that will be survive in the job market in the coming decades. This is always difficult for mentors, as we can't always anticipate the "black swan" events that might change things drastically. The tempestuous nature of her post came from one little sentence:

Contrary to what some people seem to think, R is definitely not the next big thing, either. -- AnnMaria De Mars, The Next Big Thing, AnnMaria's Blog

In SAS vs. R, Introduction and Request, Dr. Flom starts a series comparing R and SAS from the standpoint of a statistician deciding upon tools to use.

There are several threads in Dr. De Mars post. I agree with Dr. De Mars that two of the "next big things" in data management & analysis are data visualization and dealing with unstructured data. I'm of the opinion that there is a third area, related to the "Internet of Things" and the tsunami of data that will be generated by it. These are conceptual areas, however. Dr. De Mars quickly moves on to discussing the tools that might be a part of the solutions of these next big things. The concepts cited are neither software packages nor computing languages. The software packages SAS, IBM/SPSS, Stata, Pentaho and the like, and the computing language S, with its open source distribution R, and its proprietary distribution S+ are none likely to be the next big things, as they are currently useful tools to know.

I find it interesting that both Dr. De Mars and Dr. Flom, as well as the various commenters, tweeters, and other posters, are comparing software suites and applications with a computing language. I think that a bit more historical perspective might be needed in bringing these threads together.

In 1979, when I first sat down with a FORTRAN programmer to turn my Bayesian methodologies into practical applications to determine the reliability and risk associated with the STAR48 kick motor and associated Payload Assist Module (PAM), the statistical libraries for FORTRAN seemed amazing. The ease with which we were able to create the program and churn through decades of NASA data (after buying a 1MB memory box for the mainframe) was wondrous &#59;)

Today, there's not so much wonder from such a feat. The evolution of computing has drastically affected the way in which we apply mathematics and statistics today. Several of the comments to these posts argue both sides of the statement that anyone doing statistics today should be a programmer, or shouldn't. It's an interesting argument, that I've also seen reflected in chemistry, as fewer technicians are used in the lab, and the Ph.D.s work directly with the robots to prepare the samples and interpret the results.

Approximately 15 years ago, I moved from solving scientific and engineering problems directly with statistics, to solving business problems through vendor's software suites. The marketing names for this endeavor have gone through several changes: Decision Support Systems, Very Large Databases, Data Warehousing, Data Marts, Corporate Information Factory, Business Intelligence, and the like. Today, Data Mining, Data Visualization, Sentiment Analysis, "Big Data", SQL Streaming, and similar buzzwords reflect the new "big thing". Software applications, from new as well as established vendors, both open source and proprietary, are coming to the fore to handle these new areas that represent real problems.

So, one question to answer for students, is which, if any, of these software packages will best survive with and aid the growth of, their maturing careers. Will Tableau, LyzaSoft, QlikView or Viney@rd be in a better spot in 20 years, through growth or acquisition, than SAS or IBM/SPSS? Will the open source movement take down the proprietary vendors or be subsumed by them? Is Pentaho/Weka the BI & data mining solution for their career? Maybe, maybe not. But what about that other beast of which everyone speaks? Namely, R, the r-project, the R Statistical Language. What is it? Is it a worthy alternative to SAS or IBM/SPSS or Pentaho/Weka? Or is it a different genus altogether? That's a question I've been seeking to answer for myself, in my own career evolution. After 15 years, software such as SAP/Business Objects and IBM/Cognos, haven't evolved into anything that I like, with their pinnacle of statistical computation being the "average", the arithmetic mean. SAS and IBM/SPSS are certainly better, and with data mining, machine learning and predictives becoming important to business, certainly likely to be a good choice for the future. But are they really powerful enough? Are they flexible enough? Can they be used to solve the next generation of data problems?  They're very likely to evolve into software that can do so.  But how quickly?  And like all vendor software, they have limitations based upon the market studies and business decisions of the corporation.

How is R different?

Well, first, R is a computing language. Unlike SAP/Business Objects, IBM/Cognos, IBM/SPSS, SAS, Pentaho, JasperSoft, SpagoBI, or Oracle, it's not a company, nor a BI Suite, nor even a collection of software applications.  Second, R is an open source project. It's an open source implementation of S. Like C, and the other single letter named languages, S came out of Bell Labs, and in the case of R, in 1976. The open source implementation, R comes from R. Ihaka and R. Gentleman, first revealed in 1996 through the article, R: A language for data analysis and graphics. Journal of Computational and Graphical Statistics, 5:299–314, and is often associated with the Department of Statistics, University of Auckland.

While I'm not a software engineer, R is a very compelling statistical tool. As a language, it's very intuitive… for a statistician. It's an interactive, interpretive, functional, object oriented, statistical programming language. R itself is written in R, C, C++ and FORTRAN; it's powerful. As an open source project, it has attracted thousands upon thousands of users who have formed a strong community. There are thousands upon thousands of community contributed packages for R. It's flexible, and growing. One of the main goals of R was data visualization, and it has a wonderful new package for data visualization in ggplot2. It's ahead of the curve. There are packages for  parallel processing (some quite specific), for big data beyond in-memory capacity, for servers, and for embedding in a web site.  Get the idea?  If you think you need something in R, search CRAN, RForge, BioConductor or Omegahat.

As you can tell, I like R. :) However, in all honesty, I don't think that the SAS vs. R controversy is an either/or situation. SAS, IBM/SPSS and Pentaho complement R and vice-versa. Pentaho, IBM/SPSS and some SAS products support R. R can read data from SAS, IBM/SPSS , relational databases, Excel, mapReduce and more. The real question isn't is one tool better than another but rather selecting the best tool to answer a particular question.  That being said, I'm looking forward to Dr. Flom's comparison, as well as the continuing discussion on Dr. De Mars' blog.

For us, the question is building a decision support system or stack from open source components. It looks like we'll have a good time doing so.

July 2018
Mon Tue Wed Thu Fri Sat Sun
            1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31          
 << <   > >>
The TeleInterActive Press is a collection of blogs by Clarise Z. Doval Santos and Joseph A. di Paolantonio, covering the Internet of Things, Data Management and Analytics, and other topics for business and pleasure. 37.540686772871 -122.516149406889

Search

Categories

The TeleInterActive Lifestyle

Yackity Blog Blog

The Cynosural Blog

Open Source Solutions

DataArchon

The TeleInterActive Press

  XML Feeds