Machine Learning is complex for most people. One of the most common problems learners have when jumping into Machine Learning is the steep learning curve. As when you add to this the complexity of learning programming languages like Python or R you can get demotivated and lose interest fast.

But Alteryx is super powerful, the machine learning package in it enables everyone to build up a reliable ML model. Without any programming language, the only thing needed is using eyes and hands carefully.

Apart from the basic data pre-processing tools such as the formula tool, filter tool, select tool, etc… The most important tool for building the ML model is this tool:

Assisted Modelling tool, which is under the Machine Learning package tab. We can configure all the variables and parameters here without any coding. You can just drag-drop machine learning algorithms and all other functionality hiding the ugliness of code, making it much easier to grasp the fundamental concepts.

Let’s look at an example:

This is a dataset about salary, the fields of data identify 14 different conditions about a person. Based on these conditions, the person may have a salary over 50k or less than 50k.

The target of machine learning is if we know these 14 conditions of other people, can we make a prediction of whether its person’s salary is over 50k or less? Which is a typical classification model.

Here is the original data, the last column is the target, which we call the dependent variable.

  1. Using some basic Alteryx tools to clean up the data before ML. Then add an Assisted Modelling tool after it.

2. Run the workflow, now we can adjust all the parameters in the configuration window. Just click Start Assisted Modeling.

3. Select the Target of this model, which is Target, the dependent variable column. It recognises automatically, it would be a classification method.

4. Select the data type of each field, which insist of ID, Numeric and Categorical. (The ID type will be dropped as it has no effect on the result).

5. And all the way down to the last. It builds up 4 Classification models automatically and sorts them by different accuracy. Also, we can see the ROC chart and Confusion Matrix of different models.

6. Select a model you are satisfied with. Then all finished! All the tool needs to build up the model is shown in the canvas. How easy it is!

And take another test dataset to the prediction tool. That must be the easiest way to build the Machine Learning model!

 

 

Chuck Wang
Author: Chuck Wang