A Spurious Outlier Detection System for High Frequency Time Series Data

Explore a 20-minute conference talk on detecting spurious outliers in high-frequency time series data from IoT sensors. Learn about an integrated, scalable approach applicable to manufacturing, CPG, retail, healthcare, and agrotech domains. Discover how to differentiate between contextual anomalies and noisy outliers, and understand the impact of outliers on analytical models. Gain insights into the main modules of the proposed system, including thresholding, transformation, smoothing, and space filtering. Examine the workflow and second framework of this end-to-end robust system designed to improve the performance of predictive models using IoT sensor data.

Introduction
Problem Statement
Thresholding
Transformation
Transformation Thresholding
Workflow
Smoothing
Space Filtering
Second Framework