Work with pre-made Azure Machine Learning experiments to better understand the data and create predictive models.

OPTION 1 - Online Fraud Detection Guided AML Experiments

OPTION 2 - Choose your own adventure

Choose an experiment that interets you and explore the modules, trying your own data in the process.

OPTION 3 - Operationalize the AML experiment as a web service

  • Use this blog post as a guide to working with an n-gram featurizer and Kmeans clustering experiment, operationalizing it as a Python web service.

Final steps

For each of these open an intermediate dataset in a Jupyter notebook by adding an Convert to CSV module and right clicking the output node –> Open in a new Notebook and choose R or Python. Then, perform some exploratory data analysis - summarize the data, plot to understand the spread (IQRs etc.), check the variables for appropriate types (categorical vs. continous), as examples.

If you wish to Deploy as a web service follow the instructions here for an example walkthrough.