Create Your First Machine Learning Pipeline in ML.NET

[@SQLTalk]

Very easy to understand and right to the point. Thank you!

[@MyFirstReurveBow]

how is the data is already in the solution ? how is the is csv loaded in the project ? whay do you slip that ?

[@fikretakbass]

Hi Jon, many greetings from Türkiye...

[@Sam_SaaS_World]

Can you please use VS except black color and increase some zoom to easy view

[@abus3ude]

Thank you.... Can I generate label column instead of reading it from the file?

[@JonWood]

Do you mean calculate the label column instead of having it directly in the file?

[@abus3ude]

@@JonWood exactly

[@mambo7668]

I am confused which pipeline to use? The one with .Fit method or .Train ? Seems that there are 2 classes

[@fadhelalsuwaidi1221]

Hi Jon Is the test data take the whole data to test it randomly, or it takes percentage values to test for example 20 percent ?

[@JonWood]

It only takes a fraction. The newer API for it lets you pass on what fraction of data to use as test data

[@arulkumarsivaraj3517]

Where we can use pipeline algorithm? I have trained dataset and try to predict the claim prediction in simple manner but same input provide different prediction. There is any rules is existing for training data

[@JonWood]

We’re you able to evaluate the model to see how it performs?

[@arulkumarsivaraj3517]

@@JonWood same input getting different output while trying multiple times

[@BiffBifford]

Jon, I hope you are staying safe during these difficult times: I want to ask you a question? I am a trade just about every product in the stock market. I was wondering if you would consider doing a basic video on predicting stock price, nothing sophisticated but covering the basics that one might build on. The reason I ask is that computer programmers can approach the problem from a different perspective. I am unsure if you covered this subject, but I intend to review your work in greater detail. I am a "newer" sub to your content. God bless you for all your hard work in putting this rich and detailed content together for people like me to benefit. I wonder what the world will look like 1-year from today. I am not eager to find out, but it will come whether I like it or not.

[@JonWood]

There's this video I did on time series forecasting which can be something very basic you can try - kzbin.info/www/bejne/emqXh5eHodqBkLM. I know machine learning in finance is its own beast so I'll research it a bit more and try to get another video for it.

[@BiffBifford]

@@JonWood Thanks, Jon.

[@sr.programador282]

This video is gold. Thank you!

[@JonWood]

Thank you! Glad you enjoyed it!

[@sr.programador282]

​@@JonWood it was good for starting but as someone already mention they removed LearningPipeline, I am going to watch the video in the description, but thanks for this great start!

[@josbexerr5166]

Gracias Mister Jon ....Esperando mas tutox

[@reedlaww]

LearningPipeline is now under the ML.Legacy library? Is this method already deprecated?

[@JonWood]

Yep! I believe it will be eventually removed, too. You can see what the new API will look like in this video - kzbin.info/www/bejne/jXyndWWerr55osU

[@dgloeppky]

You have great tutorials, but I would be nice if you could organize them so people can progress and build on their skills in a logical order. ( from beginner to intermediate, to expert.)

[@MonkkDras]

nice article, but stack on first step. why on my ML.NET cannot find Microsoft.ML.Runtime.Api ? ML.Net version is version 1.1.0

[@MonkkDras]

Just find read the document, it was change the subject .. using Microsoft.ML.Data; class SalaryData { [LoadColumn(0)] public float YearsExperience; [LoadColumn(1), ColumnName("Label")] public float Salary; } }

[@Boaz833]

@@MonkkDras thank you so much! You're my hero!!

[@peterl1699]

Why is salary named/decorated label?

[@JonWood]

I should have mentioned that in the video. That is for ML.NET to know that that column in the training/test set is the label or column that we want to predict from the other data. Hope that helps!

[@petropzqi]

Great response. Can you have more labels from one class. I mean can we predict multiple values from the same class?

[@JonWood]

Yep, that should be possible by using a multi-label classification algorithm.

[@AAMAR0123]

Nice video... .. keep post new MI videos

[@adnen.benali]

In the video, the salaries are in a csv. How does your tutorial change when there is no csv, but the data is in a database and continues to grow?

[@JonWood]

I'm actually working on code for a video that goes over this and gets data from a database :)

[@daviddickey9832]

Much appreciated. It would be nice to also see this done with data being pulled through entity framework rather than a text file, with the data being comprised of many text columns.

[@JonWood]

Just reading through an issue in the ML.NET samples repository, it seems they don't quite have it working yet with Entity Framework. However, they are definitely working on it! (issue is here - github.com/dotnet/machinelearning-samples/issues/143#issuecomment-443306150) I'll keep an eye out and will publish a video as soon as it is released.

[@orlem5837]

Great video, but don't you think that it's not necessary to show how to do basic stuff in IDE (i.e. change platform target) in video explaining how to work with machine learning? Thank you

[@JonWood]

It was shown to show that it was required in order to use ML.NET

[@Lorenzo-kv2ql]

hi, I'm a physics student and I'm trying to learn some AI. Is this playlist about machine learning? Do you recommend watching it as an introduction to AI?

[@JonWood]

This is mostly for .NET developers looking to get into machine learning

[@MrIrrepressible]

one thing I don't understand, what is it actually training to do, what is it doing with the test data/salary data to get the final values in the console?

[@JonWood]

It's training to build a model to predict salary based on years of experience as input. To build the model it needs data to find a pattern based on the algorithm that is supplied in the pipeline. Hope that helps!

[@petropzqi]

Feel free to show us more from the ML.NEt maybe other trainers?

[@JonWood]

Have a couple more in the works :)

[@jmctx2005]

I can't Import Microsoft.ML.Runtime.Api; It returns with a red line.

[@jmctx2005]

NVM... i installed 0.3.0 and it works.

[@JonWood]

Just watch out that the API drastically changed since this video came out. Let me know if you need any guidance with it.

[@jmctx2005]

@@JonWood I tried updating to the latest ML.Net but it freaks the entire console app out. What do you use instead of API?

[@JonWood]

@@jmctx2005 Heh, yeah, they eventually removed that entire API a couple of versions ago. Try this video but an idea on the updated API - kzbin.info/www/bejne/jIfInKyLgqp-qas - but if there's still any issues let me know.

[@H8orSk8]

Im working in a Framework 4.7.2 and all the Methods are from the ML.Legacy namespace and its telling me the methods are "obsolete" any work around it?

[@JonWood]

The best thing is to see if you can convert it to the newer API. This video should help - kzbin.info/www/bejne/jXyndWWerr55osU. Btw, instead of using LocalEnvironment in that video, it is now MLContext.

[@waleedalmarshidi3494]

That was pretty easy 😶👍🏻

[@prod.dopamine]

If you guyz are having trouble with nuget package, install the version 0.3.0. Thank me later;)

[@1pavankumar23]

very nice intro project for ML. tq. can get this project download or csv files downloadable.

[@JonWood]

Glad you liked it! There is a sample project on GitHub where you can get the code and data - github.com/Wintellect/DataScienceExamples/tree/master/ML.NET

[@1pavankumar23]

@@JonWood tq . Is there any plan on creating videos on data cleaning or data pipeline creation.

[@pantepember]

@@JonWood I have been looking for the data for 10 minutes now and I've just seen this reply. It would be much better to put this link in the description area of the video. Thank you for the video, by the way.

[@JonWood]

@@pantepember Whoops! Apologies. The description has been updated with the code and dataset links. Thanks!

[@AMIRKHAN-dm2fm]

VERY NICE. PLEASE CHANGE YOU VS THEME FROM DARK TO LIGHT FOR BETTER UNDERSTANDING. THANK YOU SO MUCH