Streaming time series data to detect fraud?

Question

I'm currently looking in to the possibility of using machine learning to detect fraudulent transactions on our website based on the events that happen for each user.

I'd like to be able to stream events in to it, such as sign up, order placed, inviting another user, etc along with the the time and some how come up with a probability of how likely it is that the person is acting in a fraudulent manner

I'm totally new to machine learning and everything I read goes straight over my head basically :/

Can someone please explain the type of neural network I'd need, how I'd decide how to set it up and how I would go about training it?

Answer 1

Elasticsearch with X-pack is a tool that would come in handy. The suite has an visualization engine, Kibana and logstash that collect, parse and transform the data telemetry.

https://www.elastic.co/products/x-pack/machine-learning

Modeling can take a while to learn, luckily there are tools like Elasticsearch that can help.

Grafana also has tools but does not have a specific AI engine to it.But has a lot of integration with third party fraud detection tools.

https://grafana.com/

Sridhar