Machine learning for internet of things data analysis: a survey

Mohammad Saeid Mahdavinejad, Mohammadreza Rezvan, Mohammadamin Barekatain, Peyman Adibi, Payam M. Barnaghi, Amit P. Sheth

TL;DR

The Internet of Things (IoT) is generating massive amounts of data, and this paper surveys machine learning techniques that can effectively analyze this "smart data." It provides a taxonomy of algorithms, discusses IoT data characteristics, and highlights smart cities as a key application area, offering a valuable resource for researchers and practitioners in the field.

What Problem Does This Paper Address?

The rapid growth of the Internet of Things (IoT) has led to an explosion of data generated by interconnected devices. This "Big Data" is characterized by its high volume, velocity, and variety, posing significant challenges for traditional data analysis methods. Effectively processing and analyzing this data is crucial for developing intelligent IoT applications, especially in domains like smart cities. This paper addresses the need for a comprehensive overview of machine learning techniques suitable for IoT data, aiming to bridge the gap between the vast data generated and the intelligent insights required for smart applications.

Methods

This paper presents a survey of machine learning algorithms applicable to IoT data. The authors categorize these algorithms into eight major groups based on their structural similarities, the types of data they can handle, and their processing capabilities. They also discuss the characteristics of IoT data, particularly in the context of smart cities, and propose a taxonomy to help researchers choose appropriate algorithms. The paper delves into various machine learning paradigms, including:

  • Classification: K-Nearest Neighbors (KNN), Naive Bayes, Support Vector Machines (SVM).
  • Regression: Linear Regression, Support Vector Regression (SVR).
  • Combined Models: Classification and Regression Trees (CART), Random Forests, Bagging.
  • Clustering: K-Means, Density-Based Spatial Clustering of Applications with Noise (DBSCAN).
  • Feature Extraction: Principal Component Analysis (PCA), Canonical Correlation Analysis (CCA).
  • Neural Networks: Feed Forward Neural Networks (FFNN).
  • Time Series and Sequential Data: Algorithms for handling non-i.i.d. data.
  • Anomaly Detection: One-class Support Vector Machines (OCSVMs).

The paper also briefly touches upon computing frameworks like fog and cloud computing relevant to IoT data processing.

Datasets

Not discussed in the paper. The paper focuses on a survey of existing literature and methodologies rather than presenting new experimental results on specific datasets. However, it mentions the use of "Aarhus City smart traffic data" for a specific example of SVM application.

Key Insights and Ideas

  • Smart Data is Key: The paper emphasizes the shift from "Big Data" to "Smart Data," where the focus is on extracting actionable information and value from raw data by harnessing its volume, velocity, variety, and veracity.
  • Smart Cities as a Prime Use Case: Smart cities are highlighted as a crucial application area for IoT due to their complexity, the vast amount of data they generate, and the availability of open datasets, making them ideal for testing and developing IoT data analytics solutions.
  • Taxonomy is Essential: A clear taxonomy of machine learning algorithms is presented to guide the selection of appropriate methods for IoT data, considering its unique characteristics.
  • Data Characteristics Drive Algorithm Choice: The paper stresses that understanding IoT data characteristics (volume, velocity, variety, accuracy, dynamicity, etc.) is paramount in choosing the right machine learning algorithm for effective analysis.
  • No One-Size-Fits-All: The survey implicitly suggests that different machine learning algorithms are suited for different types of IoT data and specific tasks, underscoring the need for a nuanced approach.

Key Results

This paper is a survey and does not present novel experimental results. Its primary contribution is the comprehensive review and organization of existing knowledge. Key findings from the literature reviewed include:

  • Eight Major Algorithm Groups: The paper identifies eight major categories of machine learning algorithms applicable to IoT data, categorized by their structural similarities, data handling capabilities, and processing efficiency.
  • Smart City Data Characteristics: The paper details the characteristics of smart city data, including its continuous generation, sizable volume, varying generation rates, and the challenge of integrating heterogeneous sources.
  • Algorithm Suitability: The survey maps various machine learning algorithms to specific IoT and Smart City use cases, indicating their suitability for tasks like traffic prediction, health monitoring, energy management, and anomaly detection. For instance, SVM is noted for handling massive data and classifying diverse types, while neural networks are good for function approximation and prediction.

Limitations and Future Work

Limitations:

  • Focus on Existing Literature: As a survey, the paper's insights are limited by the scope and findings of the reviewed literature.
  • No New Datasets or Benchmarks: The paper does not introduce new datasets or benchmarks, relying on existing ones mentioned in the literature.
  • Technical Depth vs. Simplicity: While aiming for a deep and technical understanding, the paper acknowledges the need for simple implementations, suggesting a potential trade-off.

Future Work:

The paper identifies several open issues and future research directions:

  • IoT Data Characteristics: Overcoming challenges related to data quality, volume, and velocity remains a critical area. While semantic technologies can enhance data abstraction, they need further development to handle the scale and speed of IoT data.
  • IoT Applications: Addressing privacy concerns in data collection and ensuring robust security parameters for IoT applications are crucial.
  • IoT Data Analytics Algorithms: Developing algorithms that can handle real-time data from diverse sources is essential. Deep learning algorithms show promise but require sufficient data and time, and their lack of interpretability needs to be addressed. Semi-supervised algorithms are also highlighted for their potential in handling limited labeled data.

Summary updated

Abstract

Rapid developments in hardware, software, and communication technologies have allowed the emergence of Internet-connected sensory devices that provide observation and data measurement from the physical world. By 2020, it is estimated that the total number of Internet-connected devices being used will be between 25 and 50 billion. As the numbers grow and technologies become more mature, the volume of data published will increase. Internet-connected devices technology, referred to as Internet of Things (IoT), continues to extend the current Internet by providing connectivity and interaction between the physical and cyber worlds. In addition to increased volume, the IoT generates Big Data characterized by velocity in terms of time and location dependency, with a variety of multiple modalities and varying data quality. Intelligent processing and analysis of this Big Data is the key to developing smart IoT applications. This article assesses the different machine learning methods that deal with the challenges in IoT data by considering smart cities as the main use case. The key contribution of this study is presentation of a taxonomy of machine learning algorithms explaining how different techniques are applied to the data in order to extract higher level information. The potential and challenges of machine learning for IoT data analytics will also be discussed. A use case of applying Support Vector Machine (SVM) on Aarhus Smart City traffic data is presented for a more detailed exploration.
BibTeX
@inproceedings{mahdavinejad2018machine,
  author = {Mohammad Saeid Mahdavinejad and Mohammadreza Rezvan and Mohammadamin Barekatain and Peyman Adibi and Payam M. Barnaghi and Amit P. Sheth},
  title = {{Machine learning for internet of things data analysis: a survey}},
  booktitle = {Digit. Commun. Networks},
  year = {2018},
  url = {https://openreview.net/forum?id=x2tKcV3CF73},
}
Figure