How Information Systems Can Help in Alarm/Alert Detection

Florence Sedes

• Publisher: ISTE Press - Elsevier
• Year: 2018
• Print ISBN: 9781785483028
• eBook ISBN: 9780081029305
• Topic: Business
• Subtopic: Business Intelligence
• Index: Business

Alarm or alert detection remains an issue in various areas from nature, i.e. flooding, animals or earthquake, to software systems. Liveness, dynamicity, reactivity of alarm systems: how to ensure the warning information reach the right destination at the right moment and in the right location, still being relevant for the recipient, in spite of the various and successive filters of confidentiality, privacy, firewall policies, etc.? Also relevant in this context are to technical contingency issues: material failure, defect of connection, break of channels, independence of information routes and sources? Alarms with crowd media, (mis)information vs. rumours: how to make the distinction?The prediction of natural disasters (floods, avalanches, etc.), health surveillance (affectionate fevers of cattle, pollution by pesticides, etc.), air, sea and land transport, or space surveillance to prevent Risks of collisions between orbital objects involve more and more actors within Information Systems, one of whose purposes is the dissemination of alerts. By expanding the capabilities and functionality of such national or international systems, social networks are playing a growing role in dissemination and sharing, eg. with the support of systems like the Google Alert (https://www.google.fr/alerts) which concerns the publication of contents online. Recently, the Twitter microblogging platform announced a broadcast service, designed to help government organizations with alerts to the public. The proper functioning of such systems depends on fundamental properties such as resilience, liveliness and responsiveness: any alert must absolutely reach the right recipient at the right time and in the right place, while remaining relevant to him, despite the various constraints. on the one hand to external events, such as hardware failures, connection faults, breaks in communication channels, on the other hand to confidentiality, such as the collection and use of personal data (with or without the consent of the user), or the disparity of access policies (generation according to industrial, technological, security constraints, management of internal / external policies, etc.) between actors. This book opens the discussion on the "procrastination", the dynamics and the reactivity of the alert systems, but also the problems of confidentiality, filtering of information, and the means of distinguishing information and rumor.- Presents alarm or alert detection in all its aspects- Finds a solution so that the alert information reaches the right destination- Find relevance to various technical issues

1

Predicting Alarms through Big Data Analytics: Feedback from Industry Pilots

Christophe Ponsard; Annick Majchrowski; Mathieu Goeminne

Abstract

The information explosion our world is currently facing is both a challenge and opportunity for the design of information systems (IS). This chapter shows how ISs can better reach and maintain system-wide strategic goals by enabling the system to achieve predictive reasoning, which enables alarms to be raised before any negative consequences have occurred. This chapter details the use of different practical techniques from big data analytics as well as operational research on real-world examples in two different domains: IT maintenance and clinical pathways.

Keywords

Alarm terminology; Architecture; Data analytics; Goal-oriented requirements; High-level goals and KPI; Predictive maintenance; System analysis

The information explosion our world is currently facing is both a challenge and opportunity for the design of information systems (IS). This chapter shows how ISs can better reach and maintain system-wide strategic goals by enabling the system to achieve predictive reasoning, which enables alarms to be raised before any negative consequences have occurred. This chapter details the use of different practical techniques from big data analytics as well as operational research on real-world examples in two different domains: IT maintenance and clinical pathways.

1.1 Introduction

An IS is an organized system for the collection, organization, processing, storage and communication of information. It groups all of the functions (input, output, transport, processing and storage) of an application as well as databases, technical facilities and manual procedures, which support business processes [ISO 15]. Today’s information systems are present everywhere and control many personal and professional aspects of our daily lives. They are also becoming increasingly connected with the physical world due to the development of mobile devices, the emergence of the Internet of Things and cyber-physical systems.

A consequence of this increasing connectivity is the information explosion depicted in Figure 1.1. This exponential rate is widely reported as the big data area. To cite a few statistics, it is estimated that 90% of the world’s data have been produced in just the last 2 years and the amount of data created by businesses doubles every 1–2 years [ROT 15]. The zettabyte (10²¹ bytes) was reached around 2010 and by 2020, more than 40 zettabytes will be available. An important change also illustrated in the figure is that nowadays most of the data are being generated by devices rather than people.

Figure 1.1 Boom in data collection devices. For a color version of the figure, please see www.iste.co.uk/sedes/information.zip (from [MAS 16])

The main challenges an IS has to face with big data are often summarized by a series of “V” words. In addition to the Volume (i.e. the risk of information overload) already mentioned, other data dimensions are the Variety (i.e. the diversity of structured and non-structured formats), the required Velocity (i.e. highly reactive, possibly real time, data processing), the need for Visualization (in order to interpret them easily) and the related Value (in order to derive an income) [MAU 16].

An interesting opportunity raised by the volume and quality of data available is that it provides the grounds for reaching a deeper understanding of how the system actually works as well as the extent to which it fails. In this chapter, we will move away from a specific domain and use a system-level approach as defined by goal-oriented requirements engineering frameworks [VAN 01]. Figure 1.2 gives a general overview of the different types of alarms that can be raised through the combined use of monitoring of known threats identified using standard designed time techniques and runtime data analytics that are able to learn from the actual behavior.

Figure 1.2 Overview of alarm strategies based on design time analysis and runtime data analytics

Of course, reaching a higher level of anticipation of all possible problems also requires more information to be collected and the deployment of more powerful data analysis capabilities. Selecting an adequate strategy depends on many factors like cost, technical complexity and efficiency in risk reduction. It usually leads to a mixed solution where simple reactive alarms are raised for goals with limited impact while preventive alarms are generated when some known risk factors are materializing. In addition, certain predictive or even proactive analyses can be carried out to protect against the unknown, especially in systems that have less history or are very open.

This chapter aims to provide practical guidelines to make the best compromise in the design of such a strategy. Our approach will be structured as follows. As a first step, section 1.2 will give some background on the terminology, the goal level system analysis and relevant types of data analytics. It will also present the different technical options available (such as different classes of machine learning, complex event processing approaches, operation research, etc.) but without going into the implementation details. In order to provide more practical insight, section 1.3 gives an overview of our methodology and the case studies that are described in sections 1.4 and 1.5. The first study is on the maintenance of a data center in the IT domain while the second covers the management of a large set of patients engaged in chemotherapy clinical pathways in the health domain. Finally, section 1.6 provides discussions in relation to related work before we draw conclusions in section 1.7.

1.2 Background: alarm terminology, system analysis and data analytics

1.2.1 Terminology for alarm types and processing strategies

Different domains have defined their own classification of how a system can trigger alarms and react in order to ensure the continuity of system operation, for example:

• – for system maintenance, the standard ISO/IEC 16350 defines different maintenance types (illustrated in Figure 1.3) [ISO 15];

  

  
  

  Figure 1.3 Overview of ISO16350 standard for IT maintenance

• – for IT services, the Information Technology Infrastructure Library (ITIL) is widely used [OFF 11];
• – in the broader area of business continuity management, many reference frameworks have been published [HIL 10] and standardized [ISO 12], including for SMEs [ENI 10].

Unfortunately, different domains use different terms and sometimes use the same term with different meaning. In order to cope with this, we abstract away from a specific domain. Table 1.1 defines common high leve...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Introduction
• 1: Predicting Alarms through Big Data Analytics: Feedback from Industry Pilots
• 2: Mobility and Prediction: an Asset for Crisis Management
• 3: Smartphone Applications: a Means to Promote Emergency Management in France?
• 4: Mobiquitous Systems Applied to Earthquake Monitoring: the SISMAPP Project
• 5: Information Systems for Supporting Strategic Decisions and Alerts in Pharmacovigilance
• 6: An Ontologically-based Trajectory Modeling Approach for an Early Warning System
• 7: Toward a Modeling of Population Behaviors in Crisis Situations
• 8: Online Social Network Phenomena: Buzz, Rumor and Spam
• 9: How Can Computer Tools Improve Early Warnings for Wildlife Diseases?
• List of Authors
• Index