(enthusiastic techno music)
- Hello and welcome back to Microsoft Mechanics Live.
Today we're gonna take a look at the
infusion of AI and data visualization using
Power BI, the latest in enterprise scale reporting
including some things like AI generated automated insights.
First time native integrations with cognitive services
as you query your data,
new extensibility option if you're a data scientist,
and advances in reporting against big, big data sets
as well as the ability to visualize SSRS reports
within Power BI and interact with them and even more.
To do that, I'm joined by Justyna Lucznik
from the Power BI team.
Give her a big welcome.
(applause)
- Thank you.
It's good to be on the show.
- [Jeremy] Thank you.
So it's your first time on the show.
- [Justyna] Yes.
- [Jeremy] And we have an awesome audience here.
They are all using Power BI it sounded like.
So we're seeing more and more services really
across Microsoft taking in artificial intelligence
and AI and Power BI is really no exception to that, right?
- Yeah, absolutely.
So we're thinking carefully
about how we can take core Power BI capabilities,
like data visualization and data modeling,
and really combine them with AI
to create these immersive experiences.
And I really think we're taking things to the next level.
We're catering for all of our different Power BI users,
ranging from the business users, to the analysts,
and to the data scientists.
So for our business users,
as a user interacts with the report,
we're going to have machine learning
reasoning over the data and surface interesting insights.
As an analyst, you can directly apply AI enrichments
like cognitive services during your smart data prep.
And as a data scientist, with heightened integration
of product, you can now more easily use machine learning
to experiment and visualize on top of your data.
- So, again, whether you're just a Power BI user
or a data scientist, AI really becomes, in many respects,
more accessible.
And you can even--
can you walk us through actually some examples here?
- [Justyna] Yes, sure.
- [Jeremy] Everybody wants to see a demo of this.
- [Justyna] Yes, so let's start with our business user
and automated insights.
So here we have a report about Hawai'i Tourism.
And generally as a business user, what I do inside Power BI
I slice and dice my report
to try and uncover some hidden trends.
But what we're doing with Power BI
is we're trying to automate finding insights for you.
So over here I'm looking at visits by date
and I see a sort of drop in visits in U.S. West.
What I can do with Power BI
is simply right click on the visual,
select analyze, and explain the decrease.
Power BI is going to automatically find
interesting insights for me here.
- [Jeremy] And the red's probably
not a good thing there, right?
- [Justyna] No, no, those are the dips.
- [Jeremy] Okay.
- [Justyna] And over here we also as of August
are allowing you to ask questions
directly from inside the Power BI report.
So I can click on this button over here
and we automatically see some suggested questions
prepared by our analysts.
- [Jeremy] This is really cool because before
if you were a Power BI user, this is something that you did
kind of at the top level of your Power BI dashboard.
You did that from the browser as opposed to
from the desktop client.
[Justyna] Yes, so we're trying to make the
Q&A experience a lot more intuitive.
So I can, of course, type in my natural language questions.
Imagine I want to look for something like
this is my trip purpose.
- [Jeremy] Okay.
- [Justyna] Power BI is going to plot this automatically
for me.
But something new that's coming into the product
that hasn't been released yet
is this concept of asking a followup question.
So Power BI retains the context of my question
and I can ask a natural language question
or Power BI actually uses the same insights engine
to proactively prompt me with questions
that might be interesting to me.
So I can, for example, ask
what affects this distribution
and Power BI give me automatic insights.
- [Jeremy] Very cool.
So directly within Power BI you're getting kind of
machine learning, generated insights.
Even trying to predict the next question.
If you're using Excel,
and I think a lot of people probably are,
it looks a little bit similar to the ideas capability
and some of the insights that we can do in Excel
but we're actually finding data points that are interesting
as part of your data sets.
Now we saw that also demonstrated recently with
Jack Elmore when we shared all the different
AI capabilities within Office.
So how does this differ from Excel?
- [Justyna] Yes, so this uses very similar
machine learning capabilities in the backhand.
So you're going to see more and more synergy
coming between our self-service products
with Power BI, of course, being the most advanced.
[Jeremy] Okay, no bias, okay.
So you mentioned also new support for data analysts,
and there's probably a lot of data analysts out here,
so what are you doing there?
[Justyna] Yeah, so this is where we're really combining
Power BI, the power of Power BI, with the
actual ecosystem to unlock some amazing experiences.
- [Jeremy] Okay.
- [Justyna] So continuing with our Hawaii story line
over here we have the data set and the report
that's looking at hotel room use.
And you can see this Power BI report is kind of
different from usual.
It's kind of--
- [Jeremy] Unstructured-ish in terms of the data set.
- [Justyna] It is very unstructured-ish
so we have a bunch of natural kind of text fields.
Over here we have some images.
Generally in Power BI I'm used to slicing and dicing my data
so let's see how we can use--
- [Jeremy] Right.
- [Justyna] AI transforms to create some more structure
for this data.
- [Jeremy] Okay, let's do it.
- [Justyna] Yeah, so let's go ahead and navigate
to the Power BI service.
So here you see something called a data flow
and this is going to be coming soon to Power BI.
This is our self-service enterprise data prep capabilities.
And as part of a data flow,
what you can do is actually add AI insights.
And so I'm going to select this button over here.
And you can see when I click on AI insights
this is going to come up with a list of functions
that I can call,
including things like cognitive services.
- [Jeremy] And are these actually in box?
Are these something you have to add or
is this included in this function?
- [Justyna] Yeah, so the cognitive services are
all included in box.
- [Jeremy] Okay, very cool, very cool.
And the things up on the top folder,
is that something that you can add then if you've got
your own models?
- [Justyna] Yeah, so imagine you have data scientists
in your organization who have built
custom machine learning models,
you can actually bring your azure machine learning models
directly into Power BI.
- [Jeremy] Okay, cool.
So that way, if you have Python or something--
We're going to show that, I think, in a little bit
in terms of what you can do to add stuff
but how does all this work then?
How do I actually add cognitive services
to some of my reports to really put some structure
against that data?
- [Justyna] Yeah, so let's take a look.
So imagine we have hotel reviews, as we saw in the
side Power BI report
imagine I want to actually score sentiment on top of them.
All I have to do is find my text reviews field over here,
I do have to select which language I have
because this is not yet released.
In release, we'll actually detect the language.
All I have to do is invoke this function,
wait a couple of seconds,
and we're just going to create
a new calculated column for you inside this data flow
with all of your sentiment calculated
so you can see that right over here.
- [Jeremy] Very cool.
So you can tell that an analyst then
would be able to leverage all of this
without needing to understand machine learning
or ML or write any code.
These are all inbox, kind of really easy to call functions
that are just linked to various columns that you had
in your data set.
- [Justyna] Yeah, precisely.
Now imagine we wanted to use azure machine learning.
So in this particular example
my data scientist has gone ahead and built
this image classification model for doing,
image classification for hotel images.
And so I need to just select my image column over here,
again, I just want to invoke this function,
wait a couple of seconds,
and you're going to see, we basically have
a column with a bunch of URLs over here.
We can just expand this out
by selecting this button over here.
Wait a couple of seconds,
press okay, and basically we go through all the image URLs,
we score this, and you're going to see
all of the image tags appear inside my data set here.
- [Jeremy] Very cool.
So does an analyst then have to apply these functions
every single time they transform new data?
- [Justyna] No, the great thing is
these transforms just become power query steps
which all our analysts are pretty familiar with.
And so every time your data refreshes
these cognitive services and azure machine learning models
will just run on top of your data.
So, okay, we're going back to my odd, unstructured report.
And, I'm going to just go ahead--
- [Jeremy] Some interesting pictures there.
Some good and some kind of weird.
- [Justyna] Yeah, yeah.
So, you know, let's actually try and analyze these.
- [Jeremy] Okay.
- [Justyna] So I'm going to just drag over
to sentiment score over here.
And I'm gonna actually plot this by the hotel names
that I have inside my data set.
And you can very quickly see which hotels have
high sentiment and which ones are maybe not as popular.
I can click on them and cross filter my data
very easily.
- [Jeremy] Okay.
- [Justyna] But what I can also do is
bring in the image tags
that the machine learning model has found.
And so over here you can see all the different things
that the machine learning model has picked up.
- [Jeremy] And those are just sorted alphabetically now?
- [Justyna] Apparently yes.
- [Jeremy] Okay.
- [Justyna] They are sorted alphabetically
but what we can also do
is actually bring in the sentiment score
and cross correlate these together.
- [Jeremy] Okay.
- [Justyna] And now they're ranked by sentiment.
- [Jeremy] Oh, I see.
It looks like AC is at the bottom
and the beach is at the top.
That kinda makes sense.
- [Justyna] Yeah, yes.
So if you look at something like water view,
for example, you see a bunch of nice water views
but you have rightfully pointed out
AC is all the way at the bottom
and if we click on AC we see some pictures of
broken, dirty, loud ACs.
You know, I guess that makes sense.
If you're on holiday and you're taking a picture of an AC
there's probably something going on.
- [Jeremy] I think the last thing we want to do
while we're on vacation or even, you know,
while we're traveling in general, is be taking
pictures of AC units.
That means we're pretty dissatisfied with our hotel.
- [Justyna] Yes, I would say so.
And they're also mostly coming from Hotel 4.
So this is, again, kind of a question mark for me.
- [Jeremy] Right, so in this case we saw
that it actually translated
both structured and unstructured data
into a unified data set that we could actually
start to analyze and visualize using Power BI.
It's really good practical use case, in this case if you're
running a hotel, using cognitive services
directly with Power BI
and extending the learning and insights that it can do
even using that unstructured data set.
So the next level then is obviously
taking AI driven data models
and bringing them into Power BI
to further visualize that data
and actually extend the visualization capabilities
of Power BI.
What can I do there to support data scientists?
- [Justyna] Yeah, so for our data scientists
we're really looking for Power BI to be
a natural integration point with other data science tools
such as R.
And many data scientists, of course, also use Python
so as of October
we've introduced Python integration to the product, too.
So you can now use Python's grids to connect to your data.
You can plot Python visualizations
and also you can do Python power query transforms.
So everything that you can do with R,
you can now do with Python inside the desktop.
And this unlocks a ton of possibilities
ranging from creating Python machine learning models
during your data prep to
plotting Python visualizations
which don't exist in Power BI.
- [Jeremy] Right.
- [Justyna] So let me actually--
- [Jeremy] Let's see what it looks like
to use some Python in this case.
- [Justyna] Yeah, so let's flip back
to this report for a second and let's get to a new page.
In this case, I want to plot something called a swarm plot
which doesn't exist in Power BI
and I want to plot this Python script over here
which I'm just going to copy.
- [Jeremy] And of course you've got some Python handy.
- [Justyna] Yes, of course.
For the purpose of time, of course.
And so I can select this Python visualization
right over here.
All I have to do is do my data bindings
like I would do with any other Power BI visualizations.
So I'm going to select my index column over here.
I'm going to select season and I'm going to select
average temperatures
and what I want to plot is the spread of temperatures
across seasons for Hawai'i islands.
And so I'm going to go back
and select my Python visual over here and the index.
And the one other thing I'm gonna do
is make sure my data isn't summarized
because I'm really interested in plotting the spread
in this specific case.
So I'm just going to select the don't summarize buttons.
Okay, now all I have to do is copy in my script over here.
Let's make this a little bit bigger
so that we can actually see the plot
and now I just hit run.
And now Power BI is gonna go ahead
and run this Python script.
It's going to take a couple of seconds
and it's going to basically plot my swarm plot
which is going to be looking at a bunch of seasons
and the spread of temperatures.
- [Jeremy] And there it is.
- [Justyna] Yeah, there it is.
- [Jeremy] Very cool.
You can even interact with this, right?
- [Justyna] Yeah, because, you know, Power BI
is all about interactivity.
What I can do, for example, is select Maui
as the island I'm interested in
and the visual, the Python visual, will just
go ahead and cross filter.
- [Jeremy] Awesome.
So what then we can do in terms of advancing our insights
and using our data within Power BI
using machine learning,
why don't we switch gears though
and take a look at what we're doing to support
I think the next hottest topic other than AI
big data and lots of people have a lot of data
that they're trying to slice and dice through and visualize,
what are we doing there?
- [Justyna] Yeah, so with big data sets it used to be
that analysis services and reporting services were all
completely separate.
What we're doing with Power BI
is we're breaking down these silos
and we're actually bringing azure analysis services
directly into Power BI to offer a single self-service
BI platform for enterprise reporting.
And this is, I think, the biggest scalability feature
we have ever introduced to the product.
- [Jeremy] All right.
So how big does it scale?
- [Justyna] So we are going to be hopefully showing you
plotting a trillion rows of data,
which is a quarter of a petabyte of data.
- [Jeremy] Very cool.
Why don't you show us what it looks like?
And we're going to do some things with SSRS
in a bit as well, right?
- [Justyna] Yeah, so beyond scale, we're also allowing you
to very easily bring your SSRS reports
directly into the Power BI service.
And we'll do a demo of that, too.
- [Jeremy] All right, let's start with
what you're doing to support these massive data sets
and these massive scale sets.
You said a trillion rows.
- [Justyna] Yes.
- [Jeremy] So I want to see you do this
because I don't know if you can do actually
a trillion rows.
We did a billion last time
and that was already impressive.
- [Justyna] We did do a billion, yeah.
- [Jeremy] Show me a trillion.
- [Justyna] So I'm going to try to do a trillion
in five minutes.
- [Jeremy] Okay, let's try to do that.
- [Justyna] Okay.
In this particular scenario
we are looking at data coming from a smartphone app.
This is looking at a courier service and its
crowd source and so we're actually emitting
driver's locations.
And so if we look at this location count over here
this is going to count how many rows of data we have
inside this data set.
And so I'm gonna casually drag this onto the canvas.
I'm going to make this into a card.
I'm not lying to you, Jeremy.
We have a trillion rows of data here.
- [Jeremy] One trillion rows.
- [Justyna] One trillion rows, yes.
- [Jeremy] Right, and so
are we going to be able to query against this though
because it's a lot?
- [Justyna] Yeah, well.
- [Jeremy] You've got time.
- [Justyna] Let's go.
Let's give it a try.
So let's go ahead and plot something like
distance traveled.
We can plot it by activity date.
Let's go ahead and make this bigger.
Let's maybe change the visualization type.
Let's add something like miles per job
and let's actually drag miles into job into the legend.
And so you can see the query time is instantaneous.
It's like slicing and dicing through butter.
- [Jeremy] It does seem like
you're actually querying through thousands of rows maybe,
not trillions of rows.
So what's the trick to get all this to work
and be so fast in this case?
- [Justyna] Yeah, so what we're actually doing here
is we're caching the data inside Power BI like we always do
but we're actually caching the data
at the aggregated layer
and this is a new feature.
It's a new modeling feature called aggregations.
And so this allows us
to unlock these kinds of massive data sets
at essentially a fraction of the memory requirements.
- [Jeremy] Very cool.
So we're able to kind of intelligently cache that
even if you do a few steps
in terms of getting the right cache in place.
What happens then if I need to
query against stuff that might not be in the cache
that I've kind of set up as part of the aggregation step
before this?
- [Justyna] Yeah, okay.
So let's take a look.
So let's go ahead and flip this measure on.
This is showing me all the drivers
who have left the company.
And I'm going to just filter on December 23.
And what I'm going to do
is I'm going to plot a driver name over here
and I'm just going to make this table
a little bit smaller.
And I'm going to add distance traveled as well.
And so now I'm going to go and scroll through this table
and find one driver over here.
So this is Abigail Johnson.
So she left the company, she drove over 50 miles
on December 23 so I was able to kind of go down
to that level.
And now what I want to do is I want to actually
drill through on Abigail.
And I'm going to drill through
into this second report page.
And now we're no longer hitting the cache.
We have to go against the direct query basically.
And this is going against HDI Spark.
And so what you've noticed is I didn't have to go
into some other report or do anything.
It's just all---
- [Jeremy] You're not even writing an HDI Spark
or doing any of that stuff.
- [Justyna] No, and I promise you
we can go into Spark over here
and we can refresh this page over here
and we can take a quick look and see
that there's an active job that has just been kicked off.
- [Jeremy] Okay.
- [Justyna] We can go into this.
We can kind of double click
and we can see indeed we are running a query
for Abigail Johnson.
- [Jeremy] Right.
- [Justyna] And so this is actually
two amazing enterprise reporting features coming together.
The first one is composite models.
So if I hover over this table over here
you see this is a direct query table
operating against HDI Spark.
And over here, when I hover over this,
you can see that this is actually an imported table.
So we're able to now, in Power BI, bring you
connections from both import mode as well as direct query
inside one report.
- [Jeremy] And there's Abigail's travel steps
and all of the different spots where she checked in
effectively.
- [Justyna] Yes.
- [Jeremy] So very cool.
So you were able to slice into data
even if it wasn't cached locally.
It just took a couple of seconds to run.
Not too shabby in terms of going through that trillion rows.
So you mentioned that you can bring in also,
and we did mention this,
sequel server reporting services reports
also into Power BI.
I know a lot of people are using that.
In fact, I saw quite a few people here were using that.
How do we bring in existing reporting so it's
interactive in the Power BI?
- [Justyna] Yes, so over here I'm inside my Power BI
workspace like I always am.
I'm gonna navigate over to my reports
and what you'll notice is we actually have
a new type of report icon that has appeared over here.
- [Jeremy] Oh, okay.
- [Justyna] And I can drill into, let's say
my employee sales summary
and these are essentially my SSRS reports
appearing directly inside the Power BI service.
So these are pixel perfect, paginated reports.
They have interactive headers, footers, page breaks.
You can export them to PDF.
All of the features,
it's a full package, directly inside the Power BI service,
which I know a lot of people have been asking for.
- [Jeremy] So this is one of those things,
you don't have to use copy and paste
and grab like a screen snip of your SSRS report
- [Justyna] No.
- [Jeremy] and stuff it in there because
that's what a lot of people were doing that
I've heard at least.
Hopefully you don't do that.
But now you can actually get it to work and be interactive.
So it sounds like a lot of people
are actually going to be able to then
get a lot of value in this case
in terms of aggregating and unifying all that reporting
into Power BI in terms of one place to look at
both your SSRS as well as your Power BI reporting.
- [Justyna] Yeah, and they can basically collide
all their BR artifacts inside the Power BI workspace.
- [Jeremy] Very cool.
So lots of great updates.
Everybody like what you see here at Power BI?
(applause)
- [Jeremy] All right.
So we're bringing all the AI and reasoning and visualizing
over massive amounts of data as we saw here.
You can also use AI
and if you're new to Power BI you can either
start learning at Power BI.com
or even take it to the next level.
This is probably the best recommendation
in terms of where you can get started, right?
- [Justyna] Yeah, pretty much.
Definitely check out Power BI.com.
You can find video tutorials.
We encourage you to download the latest version of the
Power BI desktop
and check out the blog.
We always post new updates of all the new capabilities.
- [Jeremy] Thanks, Justyne.
And, of course, keep watching Microsoft Mechanics
as we continue to follow Power BI.
That's all the time we have for this show.
We'll see you next time.
(applause)
(enthusiastic techno music)
For more infomation >> ＃孔子最大的限制是什麼？ （感情聖化要理問答131問） - Duration: 1:32. 
For more infomation >> Kia Optima - Duration: 1:07. 
For more infomation >> Yamaha YZF R25 Yamaha R25 Facelift 2019 Tampil Agresif, Lampu LED Body Tajam - Duration: 3:38. 
For more infomation >> Kia Carens - Duration: 0:46. 
For more infomation >> Toyota Aygo 1.0-12V COMFORT 5DRS /Airco/1e Eig/NAP/Garantie - Duration: 0:46. 
For more infomation >> Volkswagen Polo 1.2 TSI Comfortline | 5drs. | navi | airco | garantie | - Duration: 1:08. 
For more infomation >> Volkswagen Polo 1.2 TSI 90pk 5drs Comfortline Executive Plus | Navigatie | LMV - Duration: 1:13. 
For more infomation >> SCUM пытаемся убить РОБОТА ! )))) - Duration: 9:00:04. 
For more infomation >> СОЛО СТРИМ,ЛБЗ 2.0 - Duration: 2:07:18. 
For more infomation >> Carros híbridos e elétricos deverão ter redução nos preços - Duration: 1:01. 
For more infomation >> Votos brancos, nulos e abstenções - Duration: 3:52. 
For more infomation >> Este é o Leite mais Saudável que Existe e 99% das Pessoas não Sabem Disso! | Naturalmente Saudável - Duration: 4:11. 
For more infomation >> NÃO É PIADA: EDUARDO BOLSONARO DIZ QUE MULHER DE DIREITA É MAIS BONITA E HIGIÊNICA - Duration: 2:08. 
For more infomation >> Tauro a quien conocerás en Octubre? Nuevo Amor + Consejo - Duration: 11:37. 
For more infomation >> O Analgésico Natural Que Funciona Como Morfina e Você Pode Encontrar no seu Quintal - Duration: 2:06. 
For more infomation >> Últimas notícia de hoje : MARCO MAIA: OS BRASILEIROS NÃO SUPORTAM A POLÍTICA DO ÓDIO - Duration: 2:34. 
For more infomation >> 영어듣기연습 #17: 빡세지만 제대로 효과있음 (딕테이션, 쉐도잉을 편리하게) - Duration: 18:52. 
For more infomation >> Mercedes-Benz E-Klasse 200 - Duration: 1:10. 
For more infomation >> Mercedes-Benz E-Klasse E 200 Coupé Automaat AMG Line - Duration: 1:09. 
For more infomation >> World's Most Expensive Shoes Cost $17 Million - Duration: 0:32. 
For more infomation >> Peugeot 308 1.2 e-THP Blue Lease Premium - Duration: 1:05. 
For more infomation >> Últimas notícia de hoje : CORREIA: A IDEOLOGIA FASCISTA DO ÓDIO SERÁ DERROTADA - Duration: 2:32. 
For more infomation >> Gott freut sich auf ein Gespräch mit dir – Joyce Meyer – Mit Jesus den Alltag meistern - Duration: 25:13. 
For more infomation >> Kavanaugh will be appointed to Supreme Court: Sen. Bill Cassidy - Duration: 3:20. 
For more infomation >> Volvo V60 2.4 D6 PiH R-DESIGN 7% BIJTELLING EXCL BTW RIJKLAAR!! - Duration: 1:14. 
For more infomation >> Volvo V60 2.0 D4 133kW/181pk Aut8 R-DESIGN BNS CLIMA + CRUISE + NAVI SENSUS + LEER/ALCANTARA + AFN.T - Duration: 1:15. 
For more infomation >> "God help the next nominee" if Kavanaugh is not confirmed, says Graham - Duration: 0:58. 
For more infomation >> Opel Mokka 1.4 T Cosmo - Duration: 1:09. 
For more infomation >> Kia cee'd 1.0 T-GDi Edition Sportswagon 120PK| Navigatie| Nieuw !!! - Duration: 0:43. 
For more infomation >> Opel Mokka 1.4 T COSMO NAVIGATIE - Duration: 1:08. 
For more infomation >> Lauren & Andy Have The Same Dream | Season 2 Ep. 2 | THE GIFTED - Duration: 1:22. 
For more infomation >> Storm Team 8 forecast: 6 p.m. 100118 - Duration: 2:48. 
For more infomation >> Superintendent won't seek second term - Duration: 0:47. 
For more infomation >> WEEKEND IN NYC VLOG | CIAO BELLA - Duration: 6:57. 
For more infomation >> Mainstream Media - RANT - Duration: 5:19. 
No comments:
Post a Comment