PFAD - Pipe flow anomaly detection

Inspiration

Our inspiration for developing our solution comes from our passion for machine learning. We saw this challenge as a way to test our new found skills from our machine learning classes. The data was not necessarily labeled, and as we have not implemented a unsupervised learning algorithm before, we saw this challenge as an opportunity to learn.

What it does

Our solution uses K-Mean clustering and custom threshold calculations to determine whether the pipe flow is abnormal or not.

How we built it

Developed AWS lambda function to take in and process pipe data from the web application, and the process results are returned to user analysis.

Challenges we ran into

• Lack of time
• Challenges with learning new technologies

Accomplishments that we're proud of

• Got the data science portion of the project working
• Working Terraform configuration to deploy application onto AWS

What we learned

• Understood how to integrate K-Means in a real world scenario
• How to use AWS Lambda functions
• Learned how to use Vue framework

What's next for PFAD - Pipe flow anomaly detection

• Handle real time information streams
• Complete front end graphics
• Use RNN on now labeled data to predict unseen datasets
• Deploy front end application onto a domain

Built With

• amazon-web-services
• python
• terrafrom
• vue