Towards an optimization model for outlier detection in IoT-enabled smart cities

EasyChair Preprint 8368

Abstract

In a connected world, the growing attention given to the IoT (Internet of Things) is driven by its economic, societal and ecological impact among others as well as its vast applications and services.

The decisions made by those applications and services are based on the data gathered from different networks of IoT sensors. A poor quality of data forwarded to control centers may lead to ill-informed decisions, inadequate services and impact adversely the business objectives.

In this paper, we will be interested in the parameters that influence the levels of data quality (DQ). These problems may be due to errors in measurements or precision of the data collection devices, energy restrictions, intermittent connectivity, interference with other devices, sampling frequency, noisy environments and data volume among others.

Data quality levels are evaluated against a set of dimensions [1]. Herein, we will focus our research on the most important dimensions for end users, such as accuracy, completeness and timeliness.

As outlier detection (OD) is a major problem in both IoT and Data Quality, it will also be addressed as a sub-domain for DQ. In fact, OD is complex since data can have a normal behavior, an outlier may be a valuable information that should not be discarded like in health diagnosis or cybersecurity.

Many techniques and methods are used for OD, each one is used for specific domains. Techniques for OD will be presented and classified. We will describe the most used techniques such as statistic-, distance-, density- clustering- and learning based methods.

A technique to detect outliers in the field of IoT-enabled smart cities will be recommended.

Keyphrases: Data Quality, Internet of Things, outlier detection

@booklet{EasyChair:8368,
  author    = {Moulay Lakbir Tahiri Alaoui and Meryam Belhiah and Soumia Ziti},
  title     = {Towards an optimization model  for outlier detection in IoT-enabled smart cities},
  howpublished = {EasyChair Preprint 8368},
  year      = {EasyChair, 2022}}