Time series analysis on AWS — Part 2 presentation — Multivariate anomaly detection
On March 4, my first book, Time Series Analysis on AWS was published! I announced it in this post and also gave some details about what you can expect in the first introductory chapter of the book. In my previous article, I also detailed the content of the next six chapters dedicated to Amazon Forecast, which is a managed service available on the AWS cloud to let you build and deploy forecasting models with limited to no knowledge of machine learning:
The last two parts of the book are dedicated to anomaly detection in different context. No one size fits all model or approach can tackle every anomaly detection use cases you might encounter. In this second part, you will learn how to leverage Amazon Lookout for Equipment to provide valuable insights to your maintenance and reliability teams. This service is geared towards multivariate time series datasets as you can find in industrial settings: sensor data, equipment data, manufacturing process data… The chapters of this part will teach you how to train and deploy your own anomaly detection model. You will also learn how to post-process these results to expand the insights you collect from the service to go beyond this and into anomaly predictions and will also give hints at how you can further go into anomaly classifications.
Note for developers and ML experts
As a developer and/or machine learning practitioner, anything you will do in the console while following along these chapters will help you understand the philosophy of the service and its limitation with regards to your datasets and use cases. Although you will likely move quickly to using the dedicated API of Amazon Lookout for Equipment (the API documentation will come in very handy then!), reading through these chapters will then help you build sound anomaly detection features you would like to integrate into your own applications.
High level overview of part 2
In the second part of the book, you will learn about Amazon Lookout for Equipment and how this AI/ML service can help you with multivariate anomaly detection. This section is structured around the following five chapters:
- Chapter 8, An overview of Amazon Lookout for Equipment: in this chapter you are going to read about the different approaches to tackle anomaly detection and learn about the specific challenges encountered with multivariate time series data. After this overview you will learn about the Amazon Lookout for Equipment service and how it can tackle part of these challenges on your behalf. Last but not least, you will also get a few pointers on when the service is likely to be well suited to your applications or not.
- Chapter 9, Creating a dataset and ingesting your data: this chapters exposes how the service expect its input data to be prepared. You will learn how to create a new project and ingest data into the service.
- Chapter 10, Training and evaluating a model: in this chapter you will train your first predictor and learn how Lookout for Equipment leverages known periods of expected anomalies to build better models. You will also get a deep dive into the evaluation dashboard and how valuable it can be to go beyond the raw results provided by the service:
- Chapter 11, Scheduling regular inferences: once you’re equipped with a trained model, you will learn how to deploy it and feed it with fresh data to collect new inference results. You will also learn how to extract these results so that you can post-process them to extract further insights of interest.
- Chapter 12, Reducing time to insights for anomaly detections: in this chapter, you will learn how to automatically process your model diagnostics using an AWS CloudWatch dashboard that will allow you to postprocess the results from a Lookout for Equipment model without writing a line a code! You will also learn how to setup an MLOps architecture to deploy an automated pipeline going from an equipment definition (list of sensors), up to an inference scheduler:
Conclusion
Upon completion of this part, you will have trained an anomaly detection model and you will know how to deploy it and collect new predictions for your fresh data. You will also have learned how to get and visualize deep insights into your model results with automated dashboards and learn how you can automate a whole pipeline to quickly scale Lookout for Equipment across multiple pieces of equipment throughout your industrial operations.
In the next post, I will present the content of the third part of the book, chapters 13 to 15, dedicated to anomaly detection and root cause analysis in business metrics with Amazon Lookout for Metrics.
The book is now available worldwide on Amazon. Here are a few links:
I hope you’ll find it interesting, feel free to leave me a comment here and don’t hesitate to subscribe to my Medium email feed if you don’t want to miss my upcoming posts! Subscription link is here:
Want to support me and future work? Join Medium with my referral link!
Let me know your thoughts!
