Anomaly Detection in real life

Course Q&A Machine Learning Specialization Advanced Learning Algorithms
1

Hello,

I recently completed the Machine Learning specialization.

I have access to quite a bit of http data in a time series format. I’m curious about setting up Anomaly Detection using this data as a bit of a fun experiment. I could start with Latency Anomaly detection perhaps and/or traffic analysis of some sort.

I found out there are a number of anomaly detection algorithms beyond what the course teaches so I’d be open to learning more in this area.

Any advice on how to get started?

What would be a good starting point?

Thanks, Matt

2

Hi @MattF that’s a great question! I would say start with a simple understanding of your data with extensive data analysis and try to identify patterns yourself, after you have a baseline build to improve the performance of a model, use ChatGPT to get basic templates and from that iterate and modify the algorithms.

I hope this helps

3

For time series data, consider a Recurrent NN. They are designed to cope with sequences, in ways that a batch process cannot.