As mentioned in Part 1, AI Builder is a game changer in Power Apps and Microsoft flow.
In this post, I am going to show how you able to access the AI Builder for classification.
to see how to set up your environment check the Part 1.
to get data, you need to click on the Data–> Entities in the right panel
then, click on the Get Data option from the top menu.
in the menu, I am choosing the Text/CSV as I already load my dataset into Blob storage.
Next, just put the link to the blob storage and click the Next, if you access via a place that needs authentication then you need to set it up here, in this scenario, I do not access a private channel.
Now, you will access the data flow editor to clean and transform your data if needed ( remove columns, combine tables…), then put a name for the data and click on the Next.
In this part, you need to adjust and map your entity, if you already have the entity, in the Load setting, you need to choose the first option ( load to the existing entity)
otherwise the second one.
then you should provide the Primary Name Field for the Titanic dataset, I choose the Name column as the passenger name. also, you may get an error regarding the change the data type to the text and you need to click on the delete rows that no longer in the query ( if you change something in the previous step), then simply click on the Next to load the data.
After loading the data, you need to refresh it, I choose to refresh it manually for the first time and click on the Creat.
Now you need to wait for data be loaded as an entity
NOTE! if you choose a binary classification, your column to predict should be in True and False format.
next, the data should be loaded and you should see a load status change to complete. otherwise, you need to back and check the data.
Now, you should see the data you loaded as a new entity in the Entity List.
Now we have data, need to create the model.
if you create the environment correctly, you should see the AI Builder in the left menu, click on the AI Builder and Build option.
Now the main page for the election of the AI model will be shown, For this scenario, we are going to choose the Binary Classification for prediction.
again be sure the column you are going to predict be in TRUE/FALSE format.
put a proper name for your model and click on the Create
Now the main process of creating the model will start.
The first step is to select which dataset need to select for training the model and what column you are going to predict.
Then, you need to choose which column of the dataset you need to be selected for training. In this example, I choose all of them which is not at all best practice but we will see how to back and change it.
Click on the next bottom and then, review what you have, if needed back and change the columns and if you happy click on the Train.
so you will see below message till the model trained
finally, you need to see the message like :
now I saw the below message, as you can see you able to check the overall accuracy, there is a training report and you able to create a New Version of this with different data, moreover in the left pannel we are not anymore in the Build section but in the Models one.
Click on the View detail in the previous page, to see the influence report.
This report shows which column has more impact on the survival of people from the Titanic ship. Also, make sure to choose the one that has made sense from a business perspective,
Name of people does not impact, also Fair and PCLASS is the same thing.
I select the Age and Pclass only. If you click on the training report, my dataset has less than 1000 rows so the report does not show here as you can see in the below picture.
At this time, to create a new training model, you need to click on New Version
This returns you to the main training page.
Now you able to select the column you want from the list.
and click on the Next to train it again.
after all experiment, now you able to publish your model to be used in a Model-driven application.
In next post I will explain how to show the results