Hello AWS community,
I'm currently exploring the AWS Lookout for Equipment service and would like to gain insights into how the service identifies the contribution of each sensor after model training. Specifically, in the inference results, such as the example below:
{"timestamp": "2024-02-02T09:26:02.000000", "prediction": 1, "prediction_reason": "ANOMALY_DETECTED", "anomaly_score": 0.95821, "diagnostics": [{"name": "scenario_4_data_experiment\\Bearing_Temp", "value": 0.19786}, {"name": "scenario_4_data_experiment\\Output_Flow", "value": 0.12428}, {"name": "scenario_4_data_experiment\\Pump_RPM", "value": 0.1344}, {"name": "scenario_4_data_experiment\\Vibration_Frequency", "value": 0.32304}, {"name": "scenario_4_data_experiment\\Vibration_RMS", "value": 0.22043}]}
I'm curious to understand how AWS Lookout for Equipment obtains the values for each individual sensor in the "diagnostics" section in result.json obtained as inference result document. What methodology or algorithm is used to attribute specific values to each sensor, and how does the service determine their contribution to the overall prediction?
Any insights or documentation pointers on this matter would be greatly appreciated.
Thank you in advance for your assistance!
Best regards