About this video
This demo video shows an example of the final application produced by three-part video series on experimenting, automating, and deploying a machine learning model using IBM® Watson™ AutoAI, learn about connecting the model API to a web app.
Transcript of this video
Basically, what I want to show you first before we get any further is kind of the complete finished product of what we’re going to do. All the documentation and everything is there online, but essentially, what we’re going to do is we’re going to build a machine learning model very quickly, within a couple minutes without knowing any data science – just using Watson’s AutoAI application. That’s going to run a bunch of algorithms and then optimize and give you eight different possible models.
We’re going to deploy one of those models and then create a web application to actually talk to our deployed machine learning model. You do not need to know any sort of data science for this, but this is kind of what it looks like. Don’t worry about the UI. I only spent a few hours on it, so it’s not very pretty or anything like that. Mostly, it’s just showing you how easy it is to use these features even if you’re not a data scientist.
What we have here is our insurance charges estimator, and you can see the estimation is made by the IBM Watson Machine Learning and Watson AutoAI. Essentially, we have a data set that has age, gender, BMI, children, smoker, and region, and it has a charge. That’s kind of how much you’re going to pay for internet charges, so this is kind of our deployed model and we picked our best one out of the eight that auto AI created. Now, we’re going to run through some different test case scenarios where we put in different fields and we try to predict that. I’ll show you kind of what is happening and also not what you want. You want to note that we’re using Watson Studio to kind of create and manage our application. AutoAI is actually building our models, and then Watson Machine Learning is taking all of these inputs and then predicting our insurance charge. That’s kind of what’s going on under the hood. We put in our age, gender, BMI, children, smoker, and region, and we click predict. That’s when we’re using our model to actually create a chart.
You can see this insurance charge is very similar to some of the data that we have in our data set, and again we got 16,000 here because we clicked smoker. You can see that once we switch from not smoker to smoker we get a much higher insurance charge, which is kind of what we would expect with insurance charges for smokers. Lastly, we’re going to change this BMI to 30, so we’re going to increase the BMI quite a bit from a healthy range to a more unhealthy range. We’ll see how that infects our insurance charges that’s predicted from our machine learning model. You can see here that the insurance prediction is now three thousand dollars more because our BMI went up, which kind of makes sense with thinking that the higher your BMI and the more unhealthy your weight-to-height ratio, the more you’re expected to pay for insurance.
This is kind of the application we’re going to build. I’ll take you step-by-step to show you how to build this, but this is kind of an educational experience. Just know that this is really for you to learn and understand how to use AutoAI and Watson Machine Learning, not for just this data set, but other data sets as well. We are doing a regression problem, but this can be done for binary classification or any other sort of classification problem.
Let’s go ahead and kind of get into the main steps of this.