All language subtitles for 05 - The Machine Learning Workflow.en

[Autogenerated] What are the steps

involved in building and training of

machine learning? Margie, let's take a

quick look in this clip off the machine

learning workload Here is what the basic

machine learning work floor looks like.

Don't be intimidated. There are lots of

processes involved here. Feel. Walk

through each of these in detail. Once

you've understood what you want to build,

you first need to look at the raw data

that you have available. What data do you

have to work with Is a sufficient to train

your machine learning? Marty? If not, you

won't be able to proceed further. You

might need to go back and seek for new

data sources. Once you know that you have

the data that you need, You can move on

and load and store the data. Get it ready

for machine learning. Make sure that it's

located in a database or a data warehouse

where you can access the data you need to

set up by planes toe extract the data from

where you have it stored. And once you

have the data with you, you need to clean

and prepare the data. This is the data pre

processing stage data in the real world

cannot be used directly to train your

machine learning models. It needs to be

pre process. Need to get rid off missing

values. Take care off out liars. If you

have no numeric representations of data,

they have to be in quarter to nomadic

form. These three processing steps that we

just discussed you when you go from raw

data toe clean data that you can feed into

a machine learning model can together be

thought off as the process of selecting

and extracting features that exist in your

data. Then you are a student off machine.

Learning your attention is mostly focused

on understanding the machine learning

algorithm and how it works. But these

three steps are critical and time

consuming steps in the real world, they

pick up an inordinate amount of time. It's

quite possible that you spend more time on

the steps than on building a machine

learning. Marty, Once you have your data

ready, the next step is to choose the

right algorithm for your use case. Do you

want a decision tree You want you support

vector machines? Do you want to use naive

bees or key nearest neighbors? The choice

of algorithm. It's up to you and dependent

on your use case. Once you've chosen an

algorithm, you'll then stream your model

on the data that you have. This is what is

referred to US fitting Ahmadi. The

training process tries to find the best

possible model parameters so that you can

use your model for prediction. Once you

have a model, you need to validate and

evaluate the model, see whether it's a

good one. There are many validation

techniques available. You'll choose a

validation method and apply the validation

method toe. Evaluate your model. You'll

examine the fit off your model and then

update the model if needed. Examining the

fit off your model is also refer to us.

According the model. You have different

metrics that you can use four different

kinds of models you have the are square

for regression models, accuracy, precision

and recall for class. If IRS, once you've

evaluated and scored, your model, will

check to see whether you're satisfied with

the result. If you're not satisfied, you

might wantto update the model. Maybe you

choose a different algorithm. Maybe you'll

use more data for training. Maybe you'll

train for longer and This is an

integrative process that continues till

you're satisfied with the model that you

have. Update the Mahdi. Choose a

validation method, examined the model

evaluated, see whether you're satisfied

and repeat till you're done. Once they're

satisfied, your model is ready to be

deployed in production and to be used for

predictions, you lose your model for

predictions on these prediction instances,

our new data points that you can then

store in your database toe, improve your

model. In the real world, prediction

instances often become partof the training

data for your Marty. This is the basic

machine learning workflow and in this

course will focus our attention on feature

selection and extraction the initial few steps.