Comments
A Chat About Big Data, Machine Learning & Value
You must login to participate in this chat. Please login.

Yes, buy the book and I'll send you a signed book plate!

Prospector

Thank you all for joining. It was great to be a guest here at A2!

Prospector

Jared. Thanks for taking the time to be with us. Great information. I'll make your final point for you: "Buy the book". Have a great day.

Editor

@MarkLR most use software. That would be my recommendation until you need to write custom things that have to execute in milliseconds then write your own.

Prospector

Have a great day everyone and thanks again Jared and Jim  ( I enjoyed the book as well ) -  Hope you can come back sometime, I know we have just skimmed the surface.

Blogger

Thanks for joining PC and Mark

Editor

Thanks to Jared and Jim for informative discussion!

Blogger

Great audio and text discussion Jim and Jared.

 

Prospector

Thanks everyone. Great info and discussions. See you next time.

 

Prospector

Enjoyed the book, talk and live chat. Thanks everyone. See you on A2

PC

Data Doctor

Jared--Do most of the organizations you work with who do "machine learning" use some analytics/statistical software or write stuff from scratch?

Prospector

Thanks, again, everyone.  Cheers...

Prospector

Folks, I know Jared has another appointment lurking. This has been a great discussion. Any final thoughts?

Editor

...until machines get a *lot* smarter, I think that is where "knowledge" (and "wisdom") will live.

Prospector

MarkLR. I know someone in K-12 who might have read something along those lines. I'll ask after she gets those brats on the bus in a few minutes.

Editor

@MarkLR - search the site for "Universities need a Data Science rethink" and other articles with the Kennesaw professor that is introduced there.

Data Doctor

@Jamescon - thanks Jim...I'll look for Beth's article. Unfortunately lots of the big data applications in education are at the post-secondary level; not so much for K-12. [Maybe that's a niche I need to carve out. :) ]

 

Prospector

But be careful:  the limit of ensemble models (hard data and gut types) is completely washed-out non-insight.  There will always be leaders with their inscrutable 'Carbon-Based' Analytics...

Prospector

Maybe empirically right Jared, but business decisions rarely have the luxury of an endless time frame. Even juries have to declare they are hung after too much analysis time.

Prospector

@Lyndon - after SF uses the model with parking for some time, they should have data that allows predictions. Also, there are toll roads in Atlanta and Dallas that use traffic dependent pricing. Seems like another opportunity for predictive modeling.

Data Doctor

@Rodney_Brown I think the right thing in a situation where gut and data disagree is to get more data and more gut (not elgently put). that is the power of ensemble models.

Prospector

@Marklr. There were some initiatives along those lines -- predicting student outcomes -- discussed here on the site last summer. Beth Schultz wrote about them, although they may have been at the college level.

Editor

@Jamescon    I am sure there is an app for that on the way !  : ) 

Blogger

@Jared    That is an interesting use of predictive analytics and machine learning.   Great question Lyndon !

Blogger

Jared writes

== I believe that the city of SF uses a machine learning model to set demand driven prices for public parking. They are striving to have 2 spots available all the time for each sector and the prices adjust based on demand.==

Not travel demand forecasting exactly, but getting closer and closer...

Blogger

Jared. Did San Francisco come up with that idea after they discovered that people were bidding for soon-to-be empty spaces?

Editor

Jared or anyone in the audience: Can you point me to some examples of where big data and analytics are being used in the K-12 Education sector, especially anything related to predictive analytics for student outcomes and instructional practices? TIA.

Prospector

@Lyndon_Henry I believe that the city of SF uses a machine learning model to set demand driven prices for public parking. They are striving to have 2 spots available all the time for each sector and the prices adjust based on demand.

Prospector

Yes, Jameson - like Information is applied data and knowledge is applied information (with experience thrown in.)  But I guess I find the terms thrown around ore as buzz words by many - so it's not always clear where the real value is under what becomes hype i guess.  And i guess that gets back to my other questions about - actually finding real business value.

Prospector

To me there are stages from transactional data to information to insights to knowledge to decisions.

Blogger

@PC    I see and I agree there is a need for "gut based" decision in the decision process  but finding the "right balance" seems to be very difficult to achieve.  

Blogger

@GaryCokins Good point! If the executives you work with don't create an enviornment where analtyics and questions are debated (at least some of the time) then I guess it is good there is a shortage and you can look for another organization to help with your talent

Prospector

I commented earlier about the tremendous utilization of analytics/big data in the public transit industry. In particular, travel demand forecasting and ridership forecasting seem classical examples of predictive analytics.

Any idea if machine learning development is impacting this implementation of predictive analytics?

 

Blogger

Jared, but if the combo of gut feeling and analytics is the perfect combo, one has to take precedent in a debate. Is data analytics the default right choice?

Prospector

@MarkLR   Good point.   Presentation is key.   Really the only chance one has in that type of environment.

Blogger

In defense of "gut feel" it is also possible for the data analysis to be incorrect. Overfitting for example. So it is a balance.

Data Doctor

@EAS4000. Good point about data, information and knowledge. Isn't knowledge just another form of data, or maybe an aggregation of data. Consider the knowledge management/transfer systems that today are capturing the knowledge of experienced field workers like repair people before they retire. Their "data" is what they know about how to fix a particular device, not just troubleshooting it, but the best way to access something like a circuit board.

Editor

Jared ... Where do the C-suite executives fit in? My observation is in the past the best leaders had the best answers. But today the best leader will have best questions. They can not rely on their past experiences and intuition that got them promoted to the executive jobs. But not all executives create a culture for questions.

Blogger

Louis: dealing with peoples' preconceived notions, especially from senior management involves "big politics". It seems no matter what the data actually shows, their minds are made up. The *presentation* of your data analysis becomes most important if you want to changes opinions.

Prospector

@Jamescon Yes for sure! But it is the combination of gut and data that makes it valuable. Those two are the ultimate ensemble model :)

Prospector

Thanks Jared ! : ) 

Blogger

@PC   Very true and probably the chief battle of Analytics promotion.

Blogger

Even after the Red Sox' success with Sabermetrics, gut feelings still rule baseball.

Prospector

@MarkLR Data and information are different things and are often treated intercahangably. Even when I'm trying not too. Information driven management is probably more correct but the risk is that information is not derived from data but from a single experience.

I don't think big information really makes sense because it is that big data that makes information and I might find a single very valuable insight from lots of data.

Prospector

LOL, nice one, @thomas314

Prospector

It's hard to change. Especially if what you've done for a long time has led to success.

Data Doctor

Anyway, thanks, All and Jared for the enlightenment...!

Prospector

Relating to Jamescon - Data - Information and knowledge seem to be increasingly sophisitcated concepts - Have people practically defined where one leaves off and the other begins?

Prospector

This gut feeling approach is something very difficult to combat, I can see many decision makers ultimately using this even after  the analyst has shown a clear case for other action.

Blogger

Well, I expect 50 years of gut/experience might be considerd 'carbon-based' Big Data.

 

Prospector

Jared.  Doesn't that "gut" feeling in business still have value if it is combined with some data as a reality check (or vice versa)?

Editor

@LouisWatson Thank you for joining. It is great to interact with an engaged community.

Prospector

A data-driven observation:  'thomas31' is pretty close to 'thomas314'!  :)

 

Prospector

Ideology has had a role in business for years. It is still difficult to convince the founder with 50-years of gut experience that he/she may be wrong.

Data Doctor

Thanks for the interesting conversation. Seems like finding the value inthe data is a big challenge.  Do companies usually have to experiment a lot to find it - or is the applied value, or monitizabel value apparant?  I see pitfalls in trying to monetize data - that usually you need to know things beyond the data you currently have - so end up trying to colect even more data before youc an monetiz...

 

Prospector

Jared--You mentioned data-driven decision making (a.k.a. "data driven management"). Do you make any distinctions between "data-driven" and "information-driven"? I know lots of folks use the terms "data" and "information" interchangably but they are different animals. Should "big data" really be "big information" ?

 

Prospector

I think there may have been one or two audience questions that we didn't get to during the audio session. If you have questions, but sure to share them with Jared now.

 

Editor

Yes I understand one of the trends is to process all data you can possibly come with for some topic.

But what about data that even when it is edited is so much that you need help sorting the ideas, or the ideas you need are somewhat hiden under a lot of text.

Prospector

Now I seem some comments/answers.  Makes sense.  Thanks.

Prospector

@thomas314 I think "gut feeling" is the business term :)

Prospector

In politics, I would guess that the analogous term to the opposite that I seek would be "ideologically-based" decision making; but business can never have worked that way, right? How did it work?

Prospector

@thomas31 - management by gut-feel. It still happens.

Data Doctor

I mentioned the discount available on Jared's book: Here are the details. Order your copy from the SAS store and get 20% off the retail price and free shipping with promo code A2BCPP. Offer ends January 31, 2015. (Store and discount are US only at this time. International community members can find purchasing options listed by country here).

Editor

@thomas314 "Hunch-driven"? "Anectdote-driven"?

Prospector

Really enjoyed the chat ....thank you Jared and Jim 

Blogger

Thank you all for joining for the chat and radio show. In addition to what Jim said about discounts on the book. If you email me or post to this chat I'll send you a signed book plate. 

Prospector

Thanks for joining, everyone.

Editor

No doubt you are right Hailey - as I said, my eye is untrained. :)

Prospector

@Rodney, i think that's true on the surface--but organizations are wanting more and more sophisticated information and extrapolation. The end result is that the job used to be arithmatic (basic queries) but now it's turning into calculus!

Prospector

Very interesting conversation, nicely done

Blogger

@Sergio, I think most companies want the data to be as raw as possible so that no possible piece of data has been cleaned out.

Prospector

As a scientist by training, I continue to be naively amazed at the degree to which business and even medicine have NOT been "data-driven"; e.g., "evidence-based medicine".  What do you call the opposite of "data-driven"?

Prospector

@kbecker - fraud is one of the examples discussed in the book. It's important to use all the data, or you may miss fraud since it's hopefully a small proportion of the data.

Data Doctor

Too true Hailey. I wonder how much or little real world experience comes in to play in dta anayltics. To my untrained eye it seems like once you learn how to structure  query, the rest is just plugging in parameters.

Prospector

I mean there are a lot of published articles on some topics like "Big Data" for instance, but you need to find a trend on a controversial issue, How about taking that kind of data? specially since Jared spoke about the limitations of some models or THE MODEL

Prospector

@Sergio, my obstacle is along that same point of having accurate and none edited data.  Our companies data is very subjectively entered (or not entered) so the end results can be manipulated - either intentionally or because of a lack of training/knowledge.  We can not wait until the data is perfect to run the analytics, so we have to make a lot of assumptions which brings that human factor right back into the analytics.  The challenge is not only to  get the leadership of the company to see this as a necessisity, but also for those involved with collecting/entering the data (if not automated) to believe that accuracy and totality is important and essential to their success too. 

Prospector

Another public agency example - tracking police cars to better cover neighborhoods and serve the public.

Data Doctor

Also public transit has been heavily using analytics & big data -- travel demand forecasting especially

Blogger

Data cleaning and integration is tough for conventional data. Does this mean that analysis of big data produces poorly integrated results?

 

Prospector

Do you see any potential to use Data Analytics to discover medicaid fraud?

Prospector

@Rodney, I hope the lag isn't too serious... experience and real-world skills take time to get!

Prospector

Hailey, from what I understand data technician or scientist is one of the fastest-growing education sectors in the past year -- right there with drone operator.

Prospector

@Sergio - published text would seem to be easier in some ways - less jargon, more context, clearer meanings. What do you mean?

Data Doctor

@Rodney, great question--and it makes me wonder if education and training has caught up with these new demands yet?

Prospector

@Jared, are those qualified people always data scientists? Or does the automation you mentioned make it easier to put non-scientists in that role?

Prospector

Any particularl challenges in getting public agencies to more effectively use big data/analytics? Are some types of agencies more amenable to using & finding benefits than others?

Blogger

What about processing data from text that has been published, I mean it has passed throug an editorial review?

Prospector

Executive sponsorship is necessary but not sufficinet.

Data Doctor

I can say that the machines at my bank have done that already. Within 4 hours of one transaction in New Hampshire, my credit card somehow bought some things at a convenience store in South Korea. No, I can't teleport.

Prospector

Howdy Hailey. Good questions.

Prospector

My question: What do you see as the biggest challenges around using technology to create an analytical environment for data mining, machine learning, and working with big data? Where are the biggest stumbling blocks?

Prospector

Speaking of privacy - how would you suggest that young people be more careful on social media?

Data Doctor

@PredictableChaos, that is true, and I have to say while they are rare, I have run across one or two very smart automated help robots.

Prospector

Via machine learning, couldn't predictive analytics be enriched to, e.g., learn lessons from customer behavior, competitor activities, etc., even current events, and recommend alternative or new strategies for the organization?

Blogger

@Rodney - yes. Businesses are happy to outsource 'chat help' windows to machines. Always cheaper than having people. Sometimes also works better.

Data Doctor

AI-level machine learning is one of the most fascinating developments in big data/analytics...

Blogger

Jared, would a business trust a mchine to learn what data is important or not?

Prospector

Yes, machine learning is often associated with AI, as an aspect of AI. Machine capability to learn...

Blogger

@Lyndon, that is my usual definition of machine learning as well.

Prospector

Machine learning is also described as enabling autonomous machines to learn from experience

Blogger

Wow. IMDB data cross-checked with Netflix data to de-anonomize individual people. That makes a lot of what happens on social media a little scary.

Data Doctor

JDCBean, can you try a forced refresh? F5 key.

Prospector

JDCBean - youhave your PC volume on, right?

 

Hello everyone

Prospector

Hmm .. can't seem to see or hear anything.  Any troubleshooting help on (at least) the audio?

Prospector

Completely agree that declining storage costs are a huge enabler. Also that it's possible to be a data 'hoarder' if there's no plan for data that's being collected and held forever.

Data Doctor

Jared, is the cost drop for data storage found in the cloud being felt for on-premises storage as well, or are the deals only found in the cloud?

Prospector

The book mentions that "no amount of data would have helped" predict the financial collapse in 2008. I understand - recent data was useless. But if we had longer term data of a different types, i think the collapse could have been predicted.

 

Data Doctor

Good morning folks.

Prospector

g'morning everyone.

Prospector

Good Afternoon Everyone!  Can't wait to hear from Mr. Dean!

Prospector

Welcome PC and Stant. 30 minute til show time.

Editor

Looking forward to the live chat.

Data Doctor

Just over an hour until today's A2 Radio Show. Let's make it a good one.

Editor


INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +