Online Program Home
My Program

Abstract Details

Activity Number: 509 - New Approaches to Modeling and Inference for Complex Space-Time Data
Type: Topic Contributed
Date/Time: Wednesday, August 1, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Physical and Engineering Sciences
Abstract #329998
Title: Automatic Anomaly Detection in Modeling Real-Time Sensor Data
Author(s): Bei Chen* and Beat Buesser
Companies: IBM Research and IBM Research
Keywords: outlier; anomaly detection ; time series ; IOT; sensor data
Abstract:

Time series collected by sensors are usually massive and noisy. One of the important steps to model such data is data cleansing. In this talk I will present two novel algorithms for anomaly detection: iterative GAM filtering and parallel recursive multi-agent search. To further improve the accuracy and flexibility, I will introduce a voting scheme for combining the results of the two algorithms. The proposed methods will be demonstrated by an example of modeling data from the sensor-equipped buildings, including energy/gas consumption, Co2 emission and occupancy.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program