Skip to content

Latest commit

 

History

History
40 lines (22 loc) · 5.63 KB

README.md

File metadata and controls

40 lines (22 loc) · 5.63 KB
Raw

Demo of IBM's Watson Studio AutoAI to Predict Titanic's survivors. BLOG HERE!

A 3-minute video of the demo can be downloaded from here. The video uses the newest version of the AutoAI UI that is not in production yet. Therefore, there will be minor chnages in the UI between the video and the demo that you can build yourself.

Here are the step-by-step instructions to build the demo on your own:

  1. If you don't an IBM cloud account, navigate to https://cloud.ibm.com/registration to create one. Otherwise, sign-in using your IBM coud account here: https://cloud.ibm.com/login

  2. Search for Watson Studio on the search bar of IBM's cloud

  3. Create a Watson Studio instance (it's a service on IBM cloud) by clicking the Create button on the bottom right

  4. Click Get Started

  5. The first step to start using WS is to create a project, click Create a project

  6. Now we want to create a project from scratch, so click Create an empty project

  7. If you already have storage, name your project and click Create (go to the next step). Otherwise, add storage by clicking Add then Create, Confirm. Then Refresh the project creation page to get your storage service attached to your project. Now click Create to create your project. A project is the key structure of Watson Studio to develop data science experiments. It is easy to add collaborators and share dataset, data connections and a big family of analytical assets (modeler flows, jupyter notebooks, data refinery flows, Watson services, among others). Feel free to spend some time exploring the different parts of a project.

  8. Let's add data to the project! Download to your laptop the public Titanic dataset from the Stanford's webpage: https://web.stanford.edu/class/archive/cs/cs109/cs109.1166/stuff/titanic.csv

  9. From your Watson Studio project, click the data icon on the top right to drop the titanic.csv or browse.

  10. Now click on Assets to verify that your data was loaded correctly.

  11. Ready for some AI? Click on Add to project blue icon on the top and click AutoAI Experiment

  12. If you already have a Watson Machine Learning (WML) service, just name your experiment and click Create (go to the next step). Otherwise, add a WML service by clicking Associate a Machine Learning instance then Create and Confirm. Now you can Reload the experiment creation page to get your WML service attached. Click Create.

  13. AutoAI only needs to be pointed to a dataset and a column to predict in the dataset. Easy. LEt's choose the titanic.csv after clicking the Select from project blue button. Click Select asset.

  14. Let's select the column to predict the titanic survivors: Survived which is a binary (takes only the values 0 = not survived or 1 = survived). The quality metric is automatically selected for you: the ROC AUC is the standard quality metric for binary classification problems. Click Run experient.

  15. The rest is waiting for AutoAI to train several pipelines and show you the best in the leaderboard. You should see the first trained pipeline in less than one minute. Once the run is completed, we can check the pipeline leaderboard to see the winner. More over, we can click the pipeline to drill down into quality metrics, confusion matrix, model information and feature importance. In this example, the most important feature to explain Titanic survivors was the sex of the passenger.

  16. To save your pipeline (and deploy as a REST API) click Save as model on the top right and Save. This action will save your model as an asset in your project.