What is AutoAI – Create and Deploy models in minutes.

What is AutoAI?

Here, we will discuss what is AutoAI and how to create and deploy any machine learning model in just a few minutes.

Developing and optimizing a machine learning model could be a time-consuming process as it requires a good amount of knowledge on models, programming language, deployment, and domain.

However, AutoAI can remove all mentioned dependency and can build an end to end Machine learning model. This also enables a non-skilled employee to build and deploy models.

AutoAI just automates the life cycle of the below processes.

• Data Load and Preparation (Pre-processing)
• Model Selection
• Feature Engineering
• Hyper-parameter optimization

Below is an example to create an AutoAI project from IBM Watson Studio. You can create a free account and try out AutoAI.

Note: At the time you are going through this, the Watson UI layout and design might change. But the overall flow will be the same.

Steps to create models in AutoAI:

Step-1: Create an account or log in to existing account. In the case of a new account, you will find like below.

 

Step-2:  Next, Create a project based on your data set, or you can use pre-existing data to create a model and play around it. We have used ‘Create a project from a sample file’.

 

Step-3: Select the tab ‘From Sample’ and you will find a number of projects listed down. Here, we have used the existing project ’Predict buying behavior with ML’.

Just note that you will be assigned a default cloud storage to use.  Select the project and click on ‘Create’.

 

Step-4: Once the project is created, you will find details of the project. Try navigating to different tabs to understand them. As a user, you can invite and give access to other members of your team.

 

 

From the ‘Asset’ tab, you can add new data files for the existing project.

 

 

Similarly, you can choose the type of running environment based on your choice from the ‘Environment’ tab.

 

 

Step-5: To understand what is AutoAI and to use AutoAI, click on the ‘Add to Project’ button and select ‘AutoAI experiment’. You will find many other asset types on this page. We will not go into other asset types as this blog focuses on AutoAI. 

 

Name the project and create it.

 

Step-6: Select data from the existing project. (As we used a sample project and the data is already available in the project)

 

Step-7: This step is one of the most important steps, where you have to define the column which you want to predict. You can choose the column which you need to predict.

-As this is a multi-class classification task, the system automatically takes ‘Prediction Type’ as ‘Multiclass Classification’ and ‘Optimized Metric’ as ‘Accuracy’.

Step-8: In case you want to change the default setting, choose the ‘Experiment setting’ button.

-This will give you a wide range of options to choose from algorithms.

-Option to choose the metric type.

-Option to choose training and testing data.

 

‘Prediction Settings’ allows users to choose different algorithms from multiple sets of algorithms. After selecting algorithms, click on ‘Save settings’.

 

Step-9: You can find the real-time progress and performance of each pipeline from the Watson dashboard. By looking at the below step, you will realize what is AutoAI and why it is called AutoAI.

 

 

The Pipeline leaderboard gives users information regarding different pipelines, its metric (accuracy ), and total build time. This also ranks different pipelines based on the result.

 

This tutorial is about What is AutoAI and an example to try out AutoAI basic features. However, there is a lot more that can be done from the Watson dashboard than what we have discussed. Give a try and happy learning.

#What is AutoAI