Sensor Dataset

A Sensor Dataset is a dataset with sensor records.

• AKA: Signal Dataset.
• Context:
  • It can (typically) include data from various Sensor Types such as temperature sensors, accelerometers, and gyroscopes.
  • It can (often) be used for real-time data analysis in applications like smart cities, health monitoring, and industrial automation.
  • It can range from being a Small Sensor Dataset to being a Large Sensor Dataset.
  • It can range from being a Clean Sensor Dataset to being a Noisy Sensor Dataset.
  • It can range from being a Static Sensor Dataset to being a Dynamic Sensor Dataset.
  • It can range from being a Single-Source Sensor Dataset to being a Multi-Source Sensor Dataset.
  • It can contain time-stamped data, making it a form of time series data.
  • It can include various types of Sensor Data quality issues such as missing data, outliers, noise, and drift.
  • It can be used to train and evaluate machine learning models for tasks like anomaly detection, predictive maintenance, and activity recognition.
  • It can be shared and accessed through data repositories and APIs for collaborative research and application development.
  • ...
• Example(s):
  • an IoT Sensor Dataset that showcases the integration of multiple sensors for smart home applications.
  • a Wearable Sensor Dataset that demonstrates the use of sensor data for health and activity monitoring.
  • ...
• Counter-Example(s):
  • Image Datasets, which contain visual data rather than sensor readings.
  • Text Datasets, which consist of textual information and are used for natural language processing tasks.
• See: Timeseries Dataset, Event Data Set, Click-Log Data Set.

References

2020

• (Teh et al., 2020) ⇒ Hui Yie Teh, Andreas W Kempa-Liehr, and Kevin I-Kai Wang. (2020). “Sensor Data Quality: A Systematic Review.” In: Journal of Big Data, 7(1).
  • NOTES:
    • Sensor data quality is crucial for the effectiveness of IoT applications, as poor data quality can render the systems useless.
    • Sensor data errors such as missing data, outliers, bias, and drift are commonly addressed by researchers to improve data quality.
    • The most common solutions for sensor data error detection are based on PCA and ANN, which together account for about 40% of all error detection methods.
    • For fault correction in sensor data, PCA, ANN, and Bayesian networks are among the most widely used techniques.
    • Sensor data errors can be corrected using methods like association rule mining, which is commonly used for imputing missing values.
    • Through systematic reviews, it has been found that methods proposed to solve sensor data errors often cannot be directly compared due to non-uniform evaluation processes and the use of non-publicly available datasets.

2004

• (Mäntyjärvi et al., 2004) ⇒ Jani Mäntyjärvi, Johan Himberg, Petri Kangas, Urpo Tuomela, and Pertti Huuskonen. (2004). "Sensor signal data set for exploring context recognition of mobile devices." In: Workshop on Benchmarks and a database for context recognition in conjunction with the 2nd International Conference on pervasive computing (PERVASIVE 2004).

• Concept
• Quality Silver