Companies ranging from energy, utilities, banking, insurance and transport gather huge amounts of time series data, such as client consumption and payment transactions. The extraction of common behaviors from these everyday life datasets can help companies improve their data quality, detect outliers and in the end, increase their financial performance.
Today’s machine learning methods enable the construction of such models. The approach of expert aggregation combined with iterative calibration clean the data by avoiding atypical points and benefit from the strengths of a multi-model approach.
For example, a fraud / demand-response detection model was set up from an aggregated load curve. The data originated from smart metering devices for fraud detection integrating a certain level of uncertainty.
Bot applications :
• Energy & Utilities - “Characterize flexibility mechanisms”
• Banking - “Detect fraudulent activities to protect both the bank and the customer”
• Transport & Industry - “Detect outliers and fraud”
• Retail - “Evaluate a common consumption behavior”