Improving Precipitation Sensor Reliability: Analysis of Four Neighbors-Based Fault Detection Methods скачать в хорошем качестве

Improving Precipitation Sensor Reliability: Analysis of Four Neighbors-Based Fault Detection Methods

KmerAI Talks#12 with Dr. Irené Tematelewo Data Scientist at Microsoft About the Guest: Irené Tematelewo is a Data Scientist at Microsoft, where he works with the Bing team to develop data-driven solutions that enhance search and user experience. He is also involved with TAHMO, the Trans-African Hydro-Meteorological Observatory weather network, where he designs, deploys and monitors quality control (QC) AI systems to promptly detect malfunctioning weather sensors, ensuring the accuracy and reliability of the collected data. Irené holds a PhD in Computer Science from Oregon State University and a Master of Engineering from Ecole Nationale Supérieure Polytechnique de Yaoundé. His doctoral research, supervised by Prof. Tom Dietterich, focused on integrated anomaly and novelty detection systems for identifying multiple types of anomalies. Topic: Improving Precipitation Sensor Reliability: Analysis of Four Neighbors-Based Fault Detection Methods Irené Tematelewo will be presenting his work entitled "Improving Precipitation Sensor Reliability: Analysis of Four Neighbors-Based Fault Detection Methods" The study is on accurate weather prediction, especially precipitation, is vital for disaster preparedness, agriculture, and resource management. Weather sensor networks provide crucial ground data for forecasting and climate studies, but sensors are prone to malfunctions and environmental interference, which can degrade data quality. This talk explores four statistical machine learning techniques that leverage neighboring sensor data to identify faulty precipitation sensors, improving data reliability. These methods are evaluated across various fault scenarios, including random faults, simultaneous faults in neighboring sensors, and faults in sensors with distant neighbors.