Predictive Maintenance in the Context of Service

A State-of-the-Art Analysis of Predictive Models and the Role of Social Media Data

in this Context

Jens Grambau

, Arno Hitzges

and Boris Otto

Faculty of Print and Media, Hochschule der Medien, Nobelstr. 10, Stuttgart, Germany

Faculty of Mechanical Engineering, TU Dortmund, Joseph-von-Fraunhofer-Str. 2-4, Dortmund, Germany

Keywords: Predictive Maintenance, Predictive Analytics, Predictive Model, Social Media, Customer Service.

Abstract: The aim of this study is to identify existing Predictive Maintenance methods in the context of service and the

role of Social Media data in this context. With the help of a Systematic Literature Review eleven researches

on notable Predictive Maintenance methods are identified and classified according to their focus, data sources,

key challenges, and assets. It can be revealed that existing methods use different Prediction technologies and

are mainly focused on industries with highly critical products. Existing methods provide value for B2B and

B2C as well as products and services. Moreover, the majority is using heterogenous data that was generated

automatically. However, it can be perceived that the consideration of Social Media data offers benefits for

Prediction methods through identifying and using personal user data, the current usage is rare and only in the

B2C sector recognizable. Thus, this research shows a gap in current literature as no universal Predictive

Maintenance solution is available, that enables organizations to enhance their services by using the full

potential of Social Media. Thus, future research needs to focus on the integration of Social Media data in

Prediction methods for the B2C sector. With this it is deeply interesting how Social Media data has to be

gathered and processed and if existing Predictive algorithms can be extended by Social Media data.

1 INTRODUCTION

Today successful companies do not only sell

excellent products, they rather offer an overall service

to customers. This service covers the customer

support along the whole product lifecycle and it

already starts with the idea of a product. Due to new

trends like Predictive Maintenance, Digital

Transformation, IoT and Big Data, companies do

have a lot of opportunities. Especially the application

of Predictive Maintenance technologies helps

companies to detect product failures and customer

dissatisfaction in an early stage. Thus, they are able

to act proactively, which leads to cost savings, the

reduction of product downtimes and an enhanced

customer service.

An essential part of Predictive Maintenance

methods and analysis consists of the collected and

analyzed amount of data. A lot of companies do not

only focus on the ongoing collection of new data, but

the use and analyses of already collected and stored

data is seen as equally important. The developments

in the information technology enable companies to

analyze data of different quantity, quality and format

and as a result economic value of this data increases

(Stockinger & Stadelmann, 2014, pp. 470–471).

Generally, the data basis can be separated in

homogenous (same data e.g. only log-data) and

heterogenous data (different data e.g. log-data and

sensor data).

With the emergence and increasing use of Social

Media networks customer or user share a lot of

personal information that had not been available

before. For companies the shared information about

products and services in Social Media represents a

new data source. The Social Media data consists of

highly specific information to single individuals as

well as to products and services.

This new data offers possibilities to enhance

Predictive methods through extending the data

sources and increasing the data quality with highly

customer specific data. Especially in the area of

customer services companies might have huge

potentials to improve profitability, customer

satisfaction, product satisfaction and lifetime of the

Grambau, J., Hitzges, A. and Otto, B.

Predictive Maintenance in the Context of Service.

DOI: 10.5220/0006669902230230

In Proceedings of the 20th International Conference on Enterprise Information Systems (ICEIS 2018), pages 223-230

ISBN: 978-989-758-298-1

223

products and thus, enhancing their customer services

through applying Prediction methods and inlcuding

highly customer specific Social Media data (Olson &

Wu, 2017, pp. 2–6).

Thus, this paper aims in providing a State-of-the-

Art analysis of existing Predictive Maintenance

methods with focus on the industries they are used in

and whether Social Media data is already considered.

First the research method is outlined through

presenting the research questions as well as the

research strategy. Second, the collected data is

presented by providing an overview of the existing

Predictive methods and approaches and their main

features. In the subsequent discussion section, the

methods are first classified and then key findings are

presented by referring back to the research questions.

Finally the main aspects are summarized in a

conclusion section and need for future research is

stated.

2 RESEARCH METHOD

2.1 Aim and Research Questions

The aim of this study is to identify existing Predictive

Maintenance methods in a service context, and the

role of Social Media data within these approaches.

To guide the ongoing literature review a series of

research questions have been developped:

1. Which basic approaches are using existing

Predictive Maintenance methods? To be

more specific, in which company sectors are

they used?

2. Are the existing approaches focused on

products or services and do they provide

value for B2B or B2C businesses?

3. Which data basis and data format is the

individual Predictive Maintenance method

using? Are existing approaches already

using Social Media data?

4. What are the challenges existing methods

are facing?

5. Which assets can be extracted from the

existing methods?

2.2 Research Strategy

A research strategy for this study had been defined,

that followed a Systematic Literature Review

procedure. Thus, the relevant data sources, time

frame and keywords for the research had been defined

in advance.

Initially a broad selection of databases for

covering diverse publications (e.g. journal articles,

conference proceedings and books) had been selected

and the search included the data sources IEEE, ACM

and JSTOR along with more traditional library

cataloguing systems. In addition, an Internet search

was conducted following a similar process for getting

a complete picture of the research area.

Moreover, keywords had been identified, that

were directly connected with Predictive Maintenance

(e.g. Predictive Analytic, Predictive Model,

Predictive Method). Many of these keywords had

been combined with ‘Service’, ‘Social’, and ‘Social

Media’ in order to ensure their relevance for this

study. The set of keywords was then adjusted after

having discovered relevant articles.

Furthermore, this study focused on literature

published in the last three years.

A large number of articles could be discovered by

browsing through the chosen databases, using the

keywords and considering the selected time period. In

a further analysis step, duplicates had been removed

and the relevance of the single articles to this study

had been checked. The abstracts of the remaining

articles had been screened and appropriate

publications had been fully read. Initially the research

identified about 100 articles, reports and books. This

had been carefully filtered to determine eleven

publications, that represent Predictive Maintenance

methods directly relevant to the research enquiry.

It is the analysis of these articles that forms the

basis of the findings in this paper.

3 EXISTING METHODS AND

APPROACHES

There has been publications of researches on

Predictive Maitenance since 1986. With the digital

transformation Predictive Maintenance gained

increased attention from researchers. Thus, there is a

significant amount of publications up from 2014 and

they mainly originate in industry nations like USA

and states of the EU. Based on the researches done in

this area, a general understanding and definitions of

the relevant terms has been established.

Thus, Predictive Maintenance can be defined as a

technique which helps to predict potential defects and

to plan maintenance intervals (Mobley, 2002, pp. 4–

6). Predictive Maintenance does not only involve the

software component - the algorithm -, it rather

consists of several components (Microsoft

Corporation, 2017). Important components are the IT

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

224

infrastructure, the used sensors for data mining, the

data itself, the availability of the data, the error-

proneness, humans as well as laws and guidelines

(Hashemian & Bean, 2011). Predictive Maintenance

refers to the idea or procedure that aims in predicting

maintenance activities (Rault & Baskiotis, 1986).

Predictive Modelling is a collection of methods to

analyze and to interpret data with the aim of deriving

predictions. These predictions are not statements

about the future, but about the probability that a

specific case will happen in a defined timeframe

(Gartner, 2017a; OnPage.org GmbH, 2017).

Predictive Analytics refers to the analysis of data, that

does not represent a prediction of the future, but

rather makes a statement about the best possible

calculated probability of a case (Gartner, 2017b;

Olson & Wu, 2017, pp. 5–6).

During the literature review eleven researches

about specific Predictive Maintenance approaches

with relevance to this study could be identified. These

existing methods vary in the industry they are focused

on and are using different technical approaches and

data. The table below summarizes main features of

each method by outlining their industry and focus,

giving a short description about their functionality,

analyzing the data sources as well as clarifying if the

method has already been validated empirically.

Table 1: Existing methods and approaches in the field of Predictive Maintenance part 1.

Method Industry/Focus Description Data sources Validation

1. Log based

maintenance

(Sipos, Fradkin,

Moerchen, &

Wang, 2014)

B2B;

Medical

industry;

Focus on

products.

Existing log data of the components as

well as information from the Service

Center is used for Predictive Maintenance

with the aim to reduce the downtime

caused by defects or to reduce

maintenance.

Preparation of big Logfiles with heavily

unstructured text. Definition of a

nomenclature for the logfiles to process

the data in a most efficient way.

Heterogenous:

Log data (which is

itself purely

homogenous text

data) and Service

Center data.

Used for several months

on a part of the device

fleet. 12 out of 31

failures could have been

detected in a one-week

prediction timeframe.

2. Bike Sharing

System

(Yang et al.,

2016)

B2C;

Transportation

industry;

Focus on

services.

The method aims in calculating and

predicting the optimized travel time, as

well as the availability of target stations

and the optimized allocation of the

bicycles’ parking stations.

A huge amount of data, fix defined

parameters (47) and weather data is

included, as it has a big impact on the

behavior of the user.

Heterogenous:

Geodata, weather

data, system data.

Tested with a data set

from the city of

Hangzhou, China and

New York City, U.S

with the same

parameters and settings.

Same performance

results of both data sets.

3. Optimizing

Life Cycle Cost

(Nichenametla,

Nandipati, &

Waghmare,

2017)

B2B;

Energy

industry;

Focus on

products.

Data from the whole product lifecycle of

Turbine blades (wind turbine) is used to

enhance service (Maintenance) and to

reach optimized maintenance costs.

All available data of the crucial

components are used as input data for the

Predictive Maintenance method.

Heterogenous:

Data from the whole

product lifecycle as

well as additional data

from suppliers and

manufactures.

Successful

implementation for a

windfarm with 300

turbine blades.

In future, the operating

costs shall be used as an

additional parameter.

4. Smart Asset

Management

(Osladil &

Kozubík, 2015)

B2B;

Energy

industry;

Focus on

products.

Method uses data about the raw material

(coal) and whether data for being able to

optimize the burning of coal through

reducing exhaust gases and thus, reducing

maintenance work during coke production.

There is a static number of parameters that

help to flexibly adjust the iterative

approach.

Heterogenous:

Sensor data, weather

data, raw material

data.

This Predictive

Maintenance method

was tested with sample

data of a single power

plant in Czech.

5. Analyzing

Healthcare Big

Data

(Sahoo,

Mohapatra, &

Wu, 2016)

B2C;

Medical

industry;

Focus on

services.

Method aims in making statements about

future health by grouping patients (high

risk, risk).

MapReduce and Intra/ Inter Data Analysis

is used to make the best possible

statements.

Sensor data (ECG, measured values) is

linked to data gained from reports (hand

written). Through the development of a

clear language the analytical procedure

can work as efficient as possible.

Heterogenous:

Data from different

sensors and machines,

hand written notes

and reports.

Very large amounts of

data.

Simulation of the

method based on

publicly available data

from clinics in

Cleveland and Hungary.

Data for 100 patients

were evaluated.

Predictive Maintenance in the Context of Service

225

Table 2: Existing methods and approaches in the field of Predictive Maintenance part 2.

Method Industry/Focus Description Data sources Validation

6. Predictive

Analytics for

Enhancing

Travel time

Pedestrian

Mode

(Amirian,

Basiri, &

Morley, 2016)

B2C;

Navigation

industry;

Focus on

services.

Method for improving a Navigation service

(pedestrian mode) through calculating the

duration for a particular route. The aim is to

increase the accuracy of the forecast by

taking additional data into account.

Use of regression models and learning

methods of the algorithms (e.g. how fast

the user usually runs).

Heterogeneous:

Personal data, weather

data, geo data, 3rd

party application data,

map data.

This Predictive

Maintenance method was

performed with 39 test

subjects. The test

subjects were located on

48 different routes,

which were used at least

five times.

7. A Social-

Network

optimized taxi

service

(Zhang, Dong,

Ota, & Guo,

2016)

B2C;

Transportation

industry;

Focus on

services.

Method aims in filling a taxi in the best

possible way with people who have a

relationship with each other (e. g. by

gender to resolve conflicts).

The approach is based on a heuristic

algorithm to find the lowest cost for each

passenger based on the length of the routes

and the customers themselves.

Linking data from social networks (e.g.

relationships between users) with data from

3rd party applications.

Heterogenous:

Social Media

relationship data, map

data, 3rd party

application data.

The method was applied

to 300 test subjects in

Hokkaido, Japan.

8. Avionic

Maintenance

Ontology based

(Palacios et al.,

2016)

B2B;

Avionic

industry;

Focus on

services.

Method aims in identifying a cause and

effect relationship for taking

countermeasures and preventing product

failure.

Ontological approach, meaning of different

language terms that mean the same.

Separation of resources (physical item) and

services (a function in the aircraft) and their

analysis for Predictive Maintenance.

Heterogeneous:

Hand-written reports,

manuals, XML, sensor

data.

Manual validation is

possible, no more

detailed description of

the performed tests.

9. A universal

sensor data

platform

modelled for

real-time

asset condition

surveillance and

big data

analytics for

railway systems

(Lee & Tso,

2016)

B2B;

Transportation

industry;

Focus on

services.

IoT approach to monitor critical parts of a

train as well as to minimize downtime and

to better plan maintenance.

Placement of the sensors on the right

components with additional analysis and

sending of data in real time.

Heterogeneous:

Sensor data from

different components

of the train.

No more detailed

description of the

validation.

10. A Pro-active

and Dynamic

Prediction

Assistance

Using

BaranC

Framework

(Hashemi &

Herbert, 2016)

B2C;

Software

industry;

Focus on

services.

A framework to predict the next user

interaction (Next App Prediction Service).

Dynamic application which is divided into

two categories: Representational (Location

and Time) and Interactional (Clicks and

Use).

The user has full control over the data

collected and shared by the BaranC

framework.

Heterogeneous:

User interactions on

the smartphone,

geodata, time data,

smartphone specific

data (network

connection, prepaid

credit, etc.).

Validation of the

Predictive model with six

test users. The test users

used about 30% of the

predicted apps in a 2-

month trial period and

the service received a

predominantly positive

feedback.

11. Social

Media analysis

on evaluating

organizational

performance: a

railway service

management

context

(Yang &

Anwar, 2016)

B2C;

Transportation

industry;

Focus on

services.

Methods aims in analyzing the influence of

Social Media on train traffic in NSW/AUS.

Evaluation of tweets to filter and measure

the performance of trains and services by

five categories (reliability, safety, security,

compactness, comfort and convenience).

The Social Media channels are used to

contact users and to publish information.

Analytical framework consists of three

modules: Data Collection, Feature

Generation, Machine Learning.

Homogenous:

Social Media data

(Twitter).

31,008 tweets from 9,428

different users were used

to validate the method.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

226

4 DISCUSSION

4.1 Classification of the Individual

Approaches

There are several works available which mention key

parameters of Predictive methods.

Each Predictive method is facing different

challenges and different factors for analyzation,

prediction and visualization need to be considered

(Chowdhury & Akram, 2011). Thus, the clear

description of the key challenges of the methods is

important (Chandola, Banerjee, & Kumar, 2009).

Rio uses a classification to describe Maintenance

methods with the help of different parameters and

asset attributes (Rio, 2017). According to Chandola

the data has to be categorized, as data is one of the

main components of Predictive methods (Chandola et

al., 2009, pp. 6–7; Han & Kamber, 2008, pp. 6–15).

Moreover, Haarman mentions the more data of

different sources is available, the better the prediction

can be done and by using available technology, as

data mining, machine learning or analytics, the

maturity level of a specific Predictive method

increases (Haarman, Mulders, & Vassiliadis, 2017,

pp. 4–9).

Thus, in order to gain a better overview of the

individual methods and for being able to compare

them, they had been classified along the following

parameters:

• ‘Focus of the Method’: This parameter will

help to see, if the existing methods are more

focused on B2C or B2B and whether they

provide value for products or services.

• ‘Data Source and Data Creation’: For this

research enquiry it is especially interesting

to see whether the approaches link different

data formats and if they are already using

Social Media data. Furthermore, it is

analyzed, if the methods use exclusively

automatically generated data or a

combination of automatically and

specifically (manually) generated data.

• ‘Challenges’: The main challenges, that

could be detected during the development of

the method and the evaluation of the

methods, are pointed out. This will help to

derive important aspects, that future

algorithms focused on Predictive

Maintenance should consider.

• ‘Assets’: Here the benefits and assets of the

specific approaches are highlighted. This is

especially interesting, as these assets might

be applied to different sectors and industries.

The following table provides a simplified

overview of the classification of the methods:

Table 3: Classification of Predictive Maintenance methods part 1.

Data Source

Method Focus

Linking of

different

Social Media

Data

Creation

Challenges Assets

1. Log based

maintenance

B2B;

Product

X Automatically

Inclusion of the data and

information from the

Service Center into the

nomenclature.

The success of a

Predictive Analytics

method must be measured

from a technical and

business point of view.

This results in different

requirements and KPI’s.

2. Bike

Sharing

System

B2C;

Service

X Automatically

Attention should be paid

to unforeseen behaviors of

the users. What happens to

the prediction if a station

is overbooked?

The prediction can be hit

in a more exact way with

the help of the huge

amount of data and the fix

parameters.

3. Optimizing

Life Cycle

Cost

B2B;

Product

X Automatically

It is important to detect

and reduce the error of

measurements of third

party data. To achieve this

a Pre-processing and Data

Reduction step is included

in the method.

Through the consideration

of the complete product

lifecycle it can be

detected, if the failure

cause lies in the

production step. This

might have an impact on a

defect or downtime of the

wind turbine.

Predictive Maintenance in the Context of Service

227

Table 4: Classification of Predictive Maintenance methods part 2.

Data Source

Method Focus

Linking of

different

Social Media

Data

Creation

Challenges Assets

4. Smart Asset

Management

B2B;

Product

X Automatically

Legal requirements for

maintenance timeframes

cannot be modified or

extended.

The use of an iterative

process in the clean-up of

the parameters and the

data provide the

possibility of an evolving

analytic method.

5. Analyzing

Healthcare

B2C;

Service

Automatically

& Specifically

Due to the use of highly

critical personal data, legal

protection is important

(agreement,

anonymization).

Merging of data generated

by machine and human

hand.

Processing of large

amounts of data.

6. Predictive

Analytics for

Enhancing

Travel Time

Pedestrian

Mode

B2C;

Service

X (X)

Automatically

& Specifically

Legal protection is

important since critical

personal data is used

(agreement,

anonymization).

The more personal data is

used, the more accurate

the prediction can be.

7. A Social-

Network

optimized taxi

service

B2C;

Service

X X

Automatically

& Specifically

The more precise the

personal data is, the better

the results can be obtained

between the users of a

specific taxi.

Linking data from social

networks with data from

third-party systems and

automatically generated

data.

8. Avionics

Maintenance

Ontology

based

B2B;

Service

Automatically

& Specifically

Redundant data must be

filtered and cleaned.

Suggestion system with a

strong focus on real-time

data to provide the best

possible Predictive

Maintenance.

9. A universal

sensor data

platform

B2B;

Service

X Automatically

Identification of sources of

interference during data

evaluation (automatic

recognition of

intermediate stops at train

stations).

Universal sensor data

platform. The detection of

component states and

planning of maintenance.

10. A Pro-

active and

Dynamic

Prediction

Assistance

B2C;

Service

Automatically

& Specifically

A challenge is the

influence of the human

psyche, especially changes

in user decisions based on

suggestions.

The learning algorithm is

based on heterogeneous

data that predicts the next

app to be used.

11. Social

Media

Analysis

B2C;

Service

X X

Automatically

& Specifically

The challenge lies in

filtering the tweets

through inclusions and

exclusions.

The approach tries to find

out what users think about

the train company. The

train company can

improve its service and

respond directly to user

criticism.

4.2 Addressing Research Questions

and Key Findings

Based on the findings of the literature review and the

classification of the relevant approaches, the research

questions stated in the beginning are addressed and

key findings are outlined.

Referring to Research Question 1, it can be

concluded that there is not one single universal

technical solution available that is best suited for

Predictive Maintenance. The analyzed methods have

selected different data mining technologies

depending on their needs and industries. Overall, all

methods attempt to derive patterns and connections

from the existing data in order to use them for

maintenance and Predictive services. Especially for

industries with highly critical products and services,

like aerospace and medical products and services,

Predictive methods have already been developed and

applied. Highly critical refers to the direct influence

on humans and their well-being.

Predictive methods have either been developed

for B2C or B2B industries and moreover, there is no

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

228

clear tendency regarding their orientation towards

product-specific solutions or services. There are

methods with product- or service-orientation. Thus,

referring to Research Question 2 the methods provide

value for B2B and B2C as well as products and

services.

Regarding the data base (Research Question 3) it

can be derived, that for the majority of the evaluated

methods, heterogeneous data that was generated

automatically, represents the data base. Nevertheless,

some methods already merge data which is generated

by humans, e.g. medical reports (see method 5) and

thus, involve subjective aspects. Three approaches

already use Social Media data for their predictions.

This data includes information about relations, posts,

tweets, comments, discussions, statuses and

reactions. Only one approach (see method 7) uses

data from social networks and combine it with data

from other sources.

One of the major challenges (Research Question

4) refers to legal aspects. It is important that the

companies inform the customers about their data

collection and make clear what kind of data is

collected and in which way. As in Europe the users

are quite skeptical about data collection (TNS

Infratest, 2016), this challenge is an important success

factor of a Predictive method (see method 4, 5 and 6).

Moreover, user's personal data highly contributes to

the quality and accuracy of the predictions. In

general, the more data is available and used,

especially personal information, the more accurate

the predictions are. Thus, the right quantity and

quality of data is highly critical for the success of

Prediction methods. For reaching this, the matching

data source for the specific prediction problem has to

be chosen and accurate filtering with inclusions and

exclusions as well as redundancy filters has to be

applied (see method 3, 7, 8, 9 and 11).

The analysis revealed (Research Question 5) that

not every Predictive Maintenance approach would

have a significant advantage by the use of Social

Media data (see method 1 and 3). Especially

approaches and methods with focus on B2B are most

likely the ones which will not benefit from Social

Media data. This is caused by the fact, that individuals

usually do not speak about highly technical B2B

products and services in Social Media and thus, in

most cases no or very little information is available.

This is confirmed by the analysis of this paper, as only

approaches with focus on B2C are currently using

Social Media data (see method 6, 7 and 11).

These methods offer new ways to identify

additional data, potential defects, and risks from a

social view, e.g. using social profile data for gaining

demographic information (see method 6), enhancing

taxi services by considering social relations (see

method 7), and improving train services through

considering customer insights and opinions about the

company (method 11). On the one hand companies

will create a closer relation to their customers

particularly in the B2C sector when considering

Social Media data. On the other hand, customers

profit from a better service, receive more information

and guidance for using a product or service in the best

way.

5 CONCLUSIONS AND FUTURE

RESEARCH

In conclusion, there are a few Predictive methods in

different industries (B2B and B2C) with focus on

service-oriented topics. For most of the existing

Predictive methods in the B2B sector, the enrichment

with Social Media data is difficult. Thus, especially

the B2C sector can profit a lot of Social Media data,

as there is much more information available which is

highly user specific and can be linked to specific

products and services.

However, only very little approaches with focus

on B2C currently link purely machine-generated data

with human-made data from social networks.

With this, a major gap in existing research is

identified, as so far there is no universal Predictive

Maintenance solution available, that helps B2C

organizations to predict and enhance their customer

services by using the full potential that can be realized

through integrating Social Media data and combine it

with other data sources.

Thus, future research should aim in answering the

following research questions:

• Which tools and keywords should be used to

gather Social Media data with focus on

products and services in the B2C sector?

• How has this data to be processed for its

further use in Predictive Maintenance

solutions? How can non-related and redundant

data be excluded by using inclusion and

exclusion parameters?

• Can algorithms of existing Predictive methods

be extended by Social Media data? How

should Social Media data be applied to?

Further research has the aim of gaining an added

value for companies and customers in the B2C sector,

by increasing the accuracy of predictions in a service

context through the consideration of Social Media

data.

Predictive Maintenance in the Context of Service

229

REFERENCES

Amirian, P., Basiri, A., & Morley, J. (2016). Predictive

Analytics for Enhancing Travel Time Estimation in

Navigation Apps of Apple, Google, and Microsoft. In:

IWCTS ’16, Proceedings of the 9th ACM SIGSPATIAL

International Workshop on Computational

Transportation Science (pp. 31–36). New York, NY,

USA: ACM. https://doi.org/10.1145/3003965.3003976

Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly

detection: A survey. ACM Comput. Surv., 41(3), 1–58.

https://doi.org/10.1145/1541880.1541882

Chowdhury, S., & Akram, A. (2011). E-Maintenance:

Opportunities and Challenges. In: TUCS Lecture Notes,

Proceedings of IRIS 2011 (pp. 68–81). Turku Centre for

Computer Science.

Gartner, I. (2017a). Predictive Analytics: IT Glossary.

Retrieved from http://www.gartner.com/it-glossary/

predictive-analytics/

Gartner, I. (2017b). Predictive Modeling: IT Glossary.

Retrieved from http://www.gartner.com/it-glossary/

predictive-modeling

Haarman, M., Mulders, M., & Vassiliadis, C. (2017).

Predictive Maintenance 4.0: Predict the unpredictable.

Retrieved from https://www.pwc.nl/en/publicaties/

predictive-maintenance-40-predict-the-

unpredictable.html

Han, J., & Kamber, M. (2008). Data mining. concepts and

techniques. Amsterdam, Heidelberg [u.a.]: Elsevier,

Morgan Kaufmann.

Hashemi, M., & Herbert, J. (2016). A Pro-active and

Dynamic Prediction Assistance Using BaranC

Framework. In: MOBILESoft ’16, Proceedings of the

International Conference on Mobile Software

Engineering and Systems (pp. 269–270). New York,

NY, USA: ACM. https://doi.org/10.1145/

2897073.2897759

Hashemian, H. M., & Bean, W. C. (2011). State-of-the-Art

Predictive Maintenance Techniques IEEE

Transactions on Instrumentation and Measurement,

60(10), 3480–3492. https://doi.org/10.1109/

TIM.2009.2036347

Lee, T., & Tso, M. (Eds.) 2016. A universal sensor data

platform modelled for realtime asset condition

surveillance and big data analytics for railway systems:

Developing a “Smart Railway” mastermind for the

betterment of reliability, availability, maintainbility

and safety of railway systems and passenger service.

2016 IEEE SENSORS.

Microsoft Corporation. (2017). Übersicht über die

vorkonfigurierte Lösung für vorhersagbaren

Wartungsbedarf. Retrieved from https://docs.

microsoft.com/de-de/azure/iot-suite/iot-suite-

predictive-overview

Mobley, R. K. (2002). An introduction to predictive

maintenance (2. ed.). Plant engineering. Amsterdam

u.a.: Butterworth-Heinemann.

Nichenametla, A. N., Nandipati, S., & Waghmare, A. L.

(Eds.) 2017. Optimizing life cycle cost of wind turbine

blades using predictive analytics in effective

maintenance planning. 2017 Annual Reliability and

Maintainability Symposium (RAMS).

Olson, D. L., & Wu, D. (2017). Predictive Data Mining

Models. Computational Risk Management. Singapore,

s.l.: Springer Singapore. Retrieved from

http://dx.doi.org/10.1007/978-981-10-2543-3

OnPage.org GmbH. (2017). Predictive Modelling.

Retrieved from https://de.onpage.org/wiki/

Predictive_Modelling

Osladil, M., & Kozubík, L. (Eds.) 2015. Smart Asset

Management in view of recent analytical technologies.

2015 16th International Scientific Conference on

Electric Power Engineering (EPE).

Palacios, L., Lortal, G., Laudy, C., Sannino, C., Simon, L.,

Fusco, G., Reynaud, C. (2016). Avionics Maintenance

Ontology Building for Failure Diagnosis Support. In:

IC3K 2016, Proceedings of the International Joint

Conference on Knowledge Discovery, Knowledge

Engineering and Knowledge Management - Volume 1

(pp. 204–209). Portugal: SCITEPRESS - Science and

Technology Publications, Lda.

https://doi.org/10.5220/0006092002040209

Rault, A., & Baskiotis, C. (Eds.) 1986. Fault detection -

Diagnosis and predictive maintenance. 1986 25th IEEE

Conference on Decision and Control.

Rio, R. (2017). IIoT Expands the Maintenance Maturity

Model. Retrieved from https://industrial-

iot.com/2017/02/iiot-expands-the-maintenance-

maturity-model/

Sahoo, P. K., Mohapatra, S. K., & Wu, S. L. (2016).

Analyzing Healthcare Big Data With Prediction for

Future Health Condition. IEEE Access, 4, 9786–9799.

https://doi.org/10.1109/ACCESS.2016.2647619

Sipos, R., Fradkin, D., Moerchen, F., & Wang, Z. (2014).

Log-based Predictive Maintenance. In: KDD ’14,

Proceedings of the 20th ACM SIGKDD International

Conference on Knowledge Discovery and Data Mining

(pp. 1867–1876). New York, NY, USA: ACM.

https://doi.org/10.1145/2623330.2623340

Stockinger, K., & Stadelmann, T. (2014). Data Science für

Lehre, Forschung und Praxis. HMD - Praxis

Wirtschaftsinform, 51(4), 469–479. https://doi.org/

10.1365/s40702-014-0040-1

TNS Infratest, M. (2016). Big Data: Wann Menschen bereit

sind, ihre Daten zu teilen. Eine europäische Studie.

Yang, J., & Anwar, A. M. (Eds.) 2016. Social Media

Analysis on Evaluating Organisational Performance a

Railway Service Management Context. 2016 IEEE 14th

Intl Conf on Dependable, Autonomic and Secure

Computing, 14th Intl Conf on Pervasive Intelligence

and Computing, 2nd Intl Conf on Big Data Intelligence

and Computing and Cyber Science and Technology

Congress(DASC/PiCom/DataCom/CyberSciTech).

Yang, Z., Hu, J., Shu, Y., Cheng, P., Chen, J., &

Moscibroda, T. (2016). Mobility Modeling and

Prediction in Bike-Sharing Systems. In : MobiSys ’16,

Proceedings of the 14th Annual International

Conference on Mobile Systems, Applications, and

Services (pp. 165–178). New York, NY, USA: ACM.

https://doi.org/10.1145/2906388.2906408

Zhang, C., Dong, M., Ota, K., & Guo, M. (2016). A Social-

Network-Optimized Taxi-Sharing Service. IT

Professional, 18(4), 34–40. https://doi.org/10.1109/

MITP.2016.71

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

230