Kickoff
Client Conversation: Saving Servers using Anomaly Detection
Import Data
Analyse Train Set to Prevent Data Snooping Bias
Birds Eye Overview
Plot Distributions
Fix Date Column to be the Appropriate Datatype
Can we visually see anomalies?
Do the spikes have anything to do with time?
Analyse Correlation
Dropping Date and Feature Engineering a Spike Column
Client Conversation: Training Anomaly Detection Models
Creating AD Pipelines
Training AD Pipelines
Client Conversation: Proxy performance
Create a Scoring Matrix for All Models
Evaluating Performance using a Proxy Metric
Making and Interpreting Predictions for Spikes
Save Models
