Author: Ajay Ohri

http://about.me/ajayohri

Louis Aslett makes data science on the cloud a 2 click step away

I was having a few issues with trying to configure the latest version of RStudio Server and the free help was not helpful enough. I came to this wonderful site and it made my job on running R on the cloud for students just a 2 click step. The best thing is lots of goodies come pre-installed.

http://www.louisaslett.com/RStudio_AMI/

Why an RStudio AMI?

The RStudio team have done a phenomenal job with making it simplicity itself to install, but there are still several motivating factors which led to me creating this AMI:

  • Although simple, it still takes several minutes to install R and RStudio after the virtual machine is going and this adds up if you do it often.
  • More time consuming is getting all the extras one may want such as LaTeX, Git, etc installed.
  • Of course, ‘simple’ is subjective and there are those who don’t know Linux, but want to use RStudio on a server without ever touching a Linux command line.
  • The EBS-backed AMIs with operating systems on tend to have vast swathes of free space which (as a postdoc of modest means) I don’t like paying to store when putting a machine into a stopped state for hibernation between computational runs! Growing an EBS volume is easier than shrinking one, so having a minimally sized AMI ready-to-go saves effort.
  • Having the full tool stack through to linking a Dropbox account in about 5 seconds means that I can go from zero to having a 36-core machine with over 200GB of RAM with all my code and data synced to a fully functional R environment with all supporting tools in a matter of minutes.
  • At the time of writing I couldn’t find any with the standard Amazon search tools and — in the great open-source tradition — that seems like an itch I should scratch!

Screenshot from 2015-09-08 13:12:48

 

Moving my Meetup Group to a Facebook Group

I am moving the 657 member Meetup Group New Delhi R Users Group

Screenshot from 2015-09-06 15:56:56to the Facebook Group New Delhi R Users Group https://www.facebook.com/groups/865780260143073/

https://www.facebook.com/groups/865780260143073/

Screenshot from 2015-09-06 15:55:17

 

 

I dont see the point of paying 30 $ every six months for an obsolete design when FB groups are more free and more easy to operate.  Privacy guidelines and spam control is just a button away on FB. LinkedIn groups are even more different.

Sponsored Post: Wise Practitioner – Predictive Analytics Interview Series: Herman Jopia of American Savings Bank

A blog post from our sponsors in the once a month series-

Original article on

http://www.predictiveanalyticsworld.com/patimes/wise-practitioner-predictive-analytics-interview-series-herman-jopia-of-american-savings-bank07152015/

In anticipation of his upcoming conference presentation,Driving Superior Growth Through Self-Developed Code, Scoring Modeling, and Price Optimization, at Predictive Analytics WorldHerman_JopiaBoston, Sept 27-Oct 1, 2015, we asked Herman Jopia, First Vice President and Data Analytics Manager at American Savings Bank, a few questions about his work in predictive analytics.

Q: In your work with predictive analytics, what behavior do your models predict?

A: We have developed and implemented attrition, profitability, and response models.

Q: How does predictive analytics deliver value at your organization? What is one specific way in which it actively drives decisions?

A: Predictive analytics helps us to understand our customers and prospects.  In practice that means a better answer to questions like who to target,  what to offer, why it makes sense, and when and how to do it.  For example, our response model for direct mail helps us to manage volume and reduce costs by excluding prospects that have a low propensity of taking the offer; therefore, it drives a lift on our profitability metrics.

Q: Can you describe a successful result, such as the predictive lift of your model or the ROI of an analytics initiative?

A: Besides monitoring the models and metrics, we actually look at how these models impact both growth and profitability.  For example in 2014 our targeted direct mail offers dramatically increased the volume and ……

Read the complete article at

http://www.predictiveanalyticsworld.com/patimes/wise-practitioner-predictive-analytics-interview-series-herman-jopia-of-american-savings-bank07152015/

#rstats Review of my book R for Cloud Computing in JSS Journal of Statistical Software

A review of R for Cloud Computing is on at Journal of Statistical Software

http://www.jstatsoft.org/v66/b04/paper

This is a lively book on a timely topic – or rather, a pair of topics, as the book is as much about R as it is on cloud computing. It should prove useful for those interested in the confluence of the two subject areas

and

The book features a number of interviews with prominent figures in data science. Though arguably a bit out of place, I believe that most readers will find them interesting and worth inclusion. This book should be of interest to anyone who is new to data storage and analysis in the cloud, especially with R, and even veteran users will find something new here and there.

and areas where the author needs to work much much harder

The book aims to provide step-by-step instructions for painlessly and quickly getting the novice user into the cloud. It does succeed in this for the most part, but any such effort will not be 100% painless after all. Readers who lack background in the cloud may feel overwhelmed at times at the beginning, given all the possible choices and myriad terms. In fact, some terms seem to be undefined, and there is no index (though there is a good bibliography). The figures are inline rather than referenced via numbers, and in some cases they are rather distant from the associated text. The font size in the figures may be too small for comfortable reading for some people.

Read the full review here http://www.jstatsoft.org/v66/b04/paper

and get a look at the full book here http://www.springer.com/book/9781493917013

 

 

Many thanks to the encouragement from Dr Matloff.

I may have been forced to drop out of U Tennessee Knoxville MS Stats on health grounds in 2010 but I get by with hard work and chutzpah.

 

Trying to improve the supply of Data Scientists without ripping young people

In a previous post, I said that many corporate are trying to benefit from the demand for data science as applied to their sector or company but not many are doing enough to improve the supply of data scientists.

demand

In anecdotal arguments for students In India and USA , many have  argued that many training companies are charging exorbitant amounts and misguided promises to essentially teach tools and techniques but not the essential analytical mindset for splicing and dicing of data as well as enough information to reach balance between the three skills for data scientists- statistics, programming and business perspective.

Added to this, many people building tools for data scientists have not worked in data science consulting them self but are addicted to one platform or product due to commercial or intellectual compulsions.

BdoZ6NjIcAAGv3w

Here is what I think could be a supply side solution to the problem of demand of data scientists hindering actual data science benefits to humanity regardless of commercial or social sectors.

  1. Build up a pool of curated best practice training
  2. Get them validated and verified across different business sectors by industry experts
  3. Add hardware or cloud training to software training
  4. Offer them on accessible platforms like mobile, tablet and web
  5. Offer them on accessible languages like Spanish Swahili Chinese Arabic as well
  6. Gamify some of the content to make it interesting, basically start creating data science hackers at an earlier age than just post graduate students
  7. Tie up with industry to offer internships that are fair balanced and demand equal commitment
  8. Tie in soft skill training for better professionalism
  9. Offer all this for free but use data generated for improving this not only on a human intervention basis but computer adaptive training and testing
  10. Monetize only after you reach a huge scale not prematurely
  11. Make it interactive using videos, 15 minute weekly personalized help on Skype from support, webinars but capture data continuously to drive engagement metrics

Do you want to just make money on the demand (uncertain) for data science but do you want to make more money on the supply side of data science too?

bor66