Strata Data Conference 2019 – San Francisco, California

About this course

then we move on to machine learning so cloud era recently merged with Hortonworks the new cloud era is twice the size of the company I sort of I'm

sure you folks are familiar with cloud era and Hort works we have discovered that we really were twins separated at birth so as we brought these two companies together one of the first

things that we needed to do was to take the sets of accounts that we both were looking at and figure out which ones were the important ones that both of our companies were focused on so we did a

bunch of different machine learning algorithms in this space because as you probably know most data sets don't have really good high quality at unique identifiers so the unique identifiers

were working with they do a lot of string distance analytics to use URLs and addresses and things like that and then the algorithms gave our sales operations folks confidence scores on

which of which set of customers should we actually all be focused on together now that we're one company and then after we brought those data sets together the next thing we need to do so

we just merge at the beginning of January now we need to organize our sales force and sort of reassign accounts to the right folks so which ones out of all those that we now have

should we really focus on in this case we use some we used our clutter data science workbench some of our underlying tools a lot of open source plug-in tools into the data science workbench to

cluster to do an unsupervised machine learning algorithm to cluster the accounts that looked alike and then we layer on top of it which via accounts have already bought and been successful

with our products and with their data and that gives us a great indication as to the other businesses out there that could be much more successful using their data from there we actually

stepped inside our accounts and we looked at where do we have opportunities that have already been identified where do we have things in the pipe and we looked at the behavior of those accounts

over time how much have they purchased from us have they used professional services are there people trained do they have strong executive champion how much are they already consuming of the

products that they've already purchased and in that place we did a lot of feature engineering and built out a model that was sort of fairly complex in nature and it brings

me to another portion of cloud era that we were able to leverage internally to make sure that we could do these algorithms much better so we have a services organization called fast

forward labs fast forward labs does two interesting things one is they advise on recommending models that you can use to produce the best results that you have

and the second thing they do is they provide a lot of research in the space of data science and Xin learning and an AI and the great thing about this is I've got a small data science team

they're super bright they're really smart but they can't possibly be connected to all of the things that are happening out in the data science in the EIA world so they leverage the advisory

services and the research from fast forward labs and then the last piece here is that we serve our analytical insights directly into some of our web applications in this case we serve a lot

of our research and our analysis on our diagnostic bundles into a web application that our support team uses and we're able to predictively open 11% of our support cases just based upon the

knowledge that we get from those models you