Automated Machine Learning or in short AutoML aims to reduce the complexity and simplify the process of training a Machine Learning Model.
In this post, we are going to show how our latest version of the platform allows you to train, integrate and deploy it as a service in this stepwise tutorial on using AutoML for regression.

AutoML Regression Process

Step 1: Deciding on the right ML task

Machine Learning models are used for different tasks, prediction, classification, and etc. In this tutorial, we want to predict a fare rate of a taxi, based on multiple parameters. The fare rate is going to be a decimal value, therefore we are going to choose a regression task in the AutoML builder.

From the Machine Learning menu, simply select the AutoML and then Regression.

Step 2: Provide the dataset

At this step, we provide the dataset source, and AutoML for regression will show us a summary of what we specified:

Step 3: Selecting columns (Features)

At this step, AutoML will require us to select input and output for the model, based on the provided data source:

Step 4: Training the model

Here is the last step of training and building the model. At this stage, everything will be automatically selected and configured for us. We just press the “Start” button to begin the process.

Our AutoML will train multiple models based on provided data for a specific amount of time, test the model automatically (cross-validation) will create everything for us.

  1. Integration with a decision model
  2. Permutation Feature Importance graph
  3. Fact concepts for data structure inputs and outputs

In this regression task using AutoML, everything is done for you like the below image illustrated in more detail:

AutoML input output process

Book a Custom Demo

First or last name is too short

In this example, we trained an ML model using AutoML for regression to predict a taxi fare amount and executed the trained ML as part of the integrated decision model with in-proc execution to predict the fare amount.

There are also, multiple deployment options are available if you need to deploy the trained model as a service. Of course, you can deploy it like any other decision model to FlexRule Server or you have multiple options to go serverless with the top 3 cloud providers; Azure Function, AWS Lambda as well as Google Cloud Function out-of-the-box.

Last updated November 23rd, 2021 at 01:25 pm, Published May 20th, 2021 at 01:25 pm