Hello and welcome my name is Manuel Quintana with Pragmatic Works, and in
collaboration with MAQ Software, we're bringing you today's video to look at
their custom visual known as Forecast using Neural Network. Now, this is a
fantastic visual, and it starts reaching into the realms of data science and
predictive analytics—which is fantastic—but it is a more advanced
area of conversation. We are gonna see how easy it is to implement this
custom visual right into your Power BI reports and start creating some
predictive analytic visuals, which is fantastic. Now, it should be noted that as
part of this visual you will need to download and install some prerequisites,
but all of this is done easily for you when you go to the Microsoft Store and
you go to install this specific visual. You'll be prompted with a message
letting you know that you do need to have some items installed, and there is a
very nice conveniently located install button right there, and it will go
through that process. What we're doing is we're installing the necessary R
packages for this visual to work. Once that's set and in place, now you can go
ahead and start using this custom visual and looking at data over time, maybe a
different type of data series, and then as well as seeing the data that you have
and you are presenting, we can now predict and get some results back from
what the neural network algorithm has learned. That is what this is all
about: leveraging the neural network learning algorithm, which is also known
like as a black box algorithm or a deep learning algorithm, and it's really great
at looking at nonlinear data. There are quite a few visuals out in the
marketplace that do predictive analytics—and they're more focused around using
algorithms that are great at handling linear data—but where neural network
shines is deriving patterns where it is nonlinear, which can be rather difficult.
Hopefully you're excited and you're ready to enjoy. Let's head over to Power
BI and see how we can put this custom visual into play.
Here we are looking at the Forecast using Neural Network from MAQ Software.
The nature of our data here is just looking at, if we look at the data view,
just the price of gold over time—so over years. Very straightforward, and that's
what we've input. If we look at the field well here for our custom visual, the
series can either be a time or numeric series and then our value. Now do note
that what we're seeing in this teal-ish color—that is our observed values. The
yellow is was what our predicted values are. So to look through this, lets go
right into the format area, so we can have an understanding of where we can
control this. The plot setting is where we can dictate the background color, the
forecast color, and the observed color, which is very important. Of course, we can
control the x and the y axis, but the other main element here is going to be
the forecast settings. Now, the default is setting this to auto, but if you have a
good understanding and you feel confident, you can actually control the
parameters that drive the neural network. We'll look at this momentarily, but I did
want to display it. Once you have this configured—and it should be noted—it
does take some time. Once you define the fields that are going to populate your
visual, the algorithm—the neural network—needs to run. It's going to be doing
analysis and creating predictions of your data. Don't worry if it's taking
a moment; that is the nature of this type of a visual. We can see that we do have
some really neat capabilities with inside of it where we can actually draw
boxes and zoom in to certain elements. By double clicking anywhere in the box, you
go back to your main view. You can enable some spike lines, so as you're moving
across, you can see how that correlates to the rest of the data—how we can see
this is a rise and there's no other peak that equals this one as far as looking
back in time. Of course, you may notice when we zoom in, in this area just
here that we have this little shaded area. This relates to the confidence
intervals. By default, this is turned off, but by turning this on, you now get a
range. You can choose—or you can see we have some confidence levels we can
choose with—lowering this number: having less confidence is gonna narrow this
range. The higher the number we're giving ourselves a little more breathing
room saying hey we're confident that the values in this time frame will fall
between these little brackets, and as you hover over,
you can see the information. The confidence level here for 2016 is gonna
be that 1373 and some change, while the yellow line is just the raw
predicted line. Really neat, really powerful, but as I mentioned, we could go
even further. If we look at the same visual but in the context of going over
to the format area and choosing to switch the forecast settings from auto
to user-defined, we have a couple of choices here. Now, of course, there's a lot
that goes into data science and learning, but we have some items here where
decay kind of controls the learning rate of the neural network. The maximum
number of iterations is how many times it's gonna run these numbers through the
neural network and coming up with different values and then coming back
with the best distribution. The number of units, in this case, is gonna represent
our series, which were going in years, so this is going to predict out to ten
years from where we ended our observed. Of course, epochs is going to be the kind
of rotations, so we're gonna have two hundred iterations over a single epoch.
Here, we're gonna have eight epochs. Lowering these values, of course, has less
iterations—there's less loading time—but there is this concept of overfitting,
which can occur if you start increasing these numbers. Basically, you're making it
learn the specifics of this data, so when new data is introduced, it may not
interpret that as correctly. Like I said, we're working in the realm of data
science, so there's a lot of information to understand here, but hopefully you can
see very quickly and how easily you can now already start using data science or
predictive model visuals here right within Power BI with the usage of this
Forecast using Neural Networks by MAQ Software. Hopefully you enjoyed, and thanks
for watching our video. If you have any questions about this visual or need a
similar business solution, feel free to contact MAQ Software at sales@maqsoftware.com.
As well, for any of your Power BI training needs, be sure to reach
out to us at Pragmatic Works by emailing training@pragmaticworks.com.
For more infomation >> Projektmanagement Software A Plan X - Neue Generation - Duration: 5:33. 
For more infomation >> Modernize your datacenter with Software-Defined Networking (SDN) in Windows Server - BRK3215 - Duration: 1:12:41. 
Không có nhận xét nào:
Đăng nhận xét