Anomaly detection and sensor analytics are used in various fields to gain competitive advantage improve service availability and customer satisfaction. Few examples are:
- Bank fraud detection
- Structural defect detection
- Medical problem detection
- Finding errors in text
- Network security and intrusion detection
- Machine failure predictions
- System health monitoring
- Event detection
Sensor analytics are used extensively to reduce cost, improve performance and customer satisfaction. Following are few examples of industry use:
- Fleet management: Sensor data from delivery trucks is helping businesses schedule preventive maintenance before mechanical issues can disrupt fleet operations.
- Healthcare sensing: Biosensors are now used to enable better and more efficient patient care across a wide range of healthcare operations, including telemedicine, telehealth, and mobile health.
- Product monitoring: Manufacturers use sensor-data analytics to monitor the health and performance of their products and to work proactively to address service and maintenance issues before they lead to product downtime.
- Predictive maintenance: Manufacturing companies use data from various sensors from their products to perform predictive maintenance before the equipment breakdown to save expensive repairs and loss of service.
- Smart buildings: Real estate management companies use building data such has HVAC data, ventilation etc. to do preventive maintenance and save expensive repairs and loss of service.
- Smart drilling: Oil and gas industry uses sensor technology to predict probability of finding oil at certain location and reduce experimental drilling costs.
- Smart grids: Governments, local cities are upgrading the grid based on data received from smart grids by predicting future usage and capacity.
- Smart meters: Utility companies can gather data from the smart meter to model and predict usage patterns of various consumers to gain competitive advantage.
- Usage based insurance: Insurance companies can use sensor data to predict probability of a claim on driver’s driving profile. This helps insurance company adjust the premium so that they can gain competitive advantage by quoting cheaper rates to safe drivers and charging extra premium to risky drivers.
For anomaly detection we use various techniques, here are a few that we have used:
- Supervised learning (Neural networks, support vector machines, random forest algorithms)
- Unsupervised learning (multivariate normal modelling, clustering techniques)
- Semi-supervised techniques
- Density based techniques such as K-nearest neighbor, local outlier factor
- Fuzzy logic based outlier detection
- Ensemble techniques possibly combining multiple techniques.