In current research, scientists are reporting on the support of automated production planning and control systems. As the latest findings allow planners to put their ideas into practice, the focus continually moves to new challenges. For example, the planning of a new production facility involves the manual design of several alternative models, which are then tested and optimized using virtual simulation – a time‐consuming and costly procedure [29].
A common problem encountered by many pattern recognition processes is incomplete or missing data. The ability to process incomplete data is a fundamental prerequisite for pattern classification, since the omission of certain values for relevant data attributes can seriously compromise the accuracy of the classification results [30].
30.8 Conclusion and Outlook
When looking at developments in machine learning over the last few decades, it is evident that the terminology and methods in this context have evolved on both technological as well as algorithmic and functional levels. If we consider that the complex issues of machine learning and predictive analytics are supported by highly parametric, multi‐variant mathematical functions, which can be trained or applied to new problems using sophisticated methods such as neural networks or deep learning, it becomes clear that the rapid growth of all recent developments is, on the one hand, thanks to the major developments in hardware and, on the other, to the high scalability made possible by cloud solutions.
Although this may make us think that developments could go on growing in this direction forever, we should always be aware that technology faces physical limitations as well.
So it could be that the next era of machine learning is not ushered in by taking another step towards more technological development, like we have seen with In‐Memory‐Computing or Microsoft Azure ML in recent years. Instead the next step might be something more revolutionary, such as new insights into genetics and evolution and a deeper understanding of the human brain above and beyond classic neural networks/deep learning, which can then be transferred to deal with technological challenges. These new prospects for research will require adaptations to existing algorithms and, much more than that, the creation of new ones.
References
1.
Y. D. W. &. C. Q. Yu, The impact of social and conventional media on firm equity value: A sentiment analysis approach, Decision Support Systems, 55.
2.
I. Witten, E. Frank and M. Hall, Data Mining: Practical Machine Learning Tools and Techniques, Burlington: Morgan Kaufmann Publishers, 2011.
3.
WinfWiki, “Analyse Text Mining mit R,” 2014. [Online]. Available: http://winfwiki.wi-fom.de/images/5/5d/A_Venn_Diagram_of_the_Intersection_of_text_mining.PNG. [Accessed 14 08 2016].
4.
A. Sharafi, Knowledge Discovery in Databases. Eine Analyse des Änderungsmanagements in der Produktentwicklung, Wiesbaden: Springer Gabler, 2012.
5.
F. Pimenta, D. Obradovic, R. Schirru and S. &. D. A. Baumann, Automatic Sentiment, 2010.
6.
B. Pang and L. &. V. S. Lee, “Thumbs up? Sentiment classification using machine learning techniques,” in Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2002.
7.
F. Nielsen and M. &. H. L. Etter, Real-time monitoring of sentiment in business related Wikipedia articles, 2013.
8.
P. &. L. M. Turney, “Measuring praise and criticism: Inference of semantic orientation from association,” in ACM Transactions on Information Systems, 2003.
9.
C. Manzei, L. Schleupner and R. Heinze, Industrie 4.0 im internationalen Kontext. Kernkonzepte, Ergebnisse, Trends, Berlin: VDE VERLAG GMBH, 2016.
10.
G. Webb, “Naive Bayes,” in Encyclopedia of Machine Learning, Springer Verlag, 2010, p. 713.
11.
C. R. P. &. S. H. Manning, Introduction to Information Retrieval, Cambridge University Press, 2008.CrossrefMATH
12.
X. Zhang, “Support Vector Machines,” in Encyclopedia of Machine Learning, Springer Verlag, 2010, pp. 941–946.Crossref
13.
D. Jansen, Sentiment Classification, Duisburg: Universität Duisburg-Essen, 2008.
14.
A. Ratnaparkhi, “Maximum Entropy Models,” in Encyclopedia of Machine Learning, Springer Verlag, 2010, pp. 647–651.
15.
M. Schoenherr, “Ohne Verstand ans Ziel,” 2016. [Online]. Available: http://www.deutschlandfunk.de/maschinelles-lernen-ohne-verstand-ans-ziel.740.de.html?dram:article_id=349980. [Accessed 08 01 2016].
16.
L. A. Zadeh, Fuzzy Sets, Information and Control, Elsevier Inc., 1965.MATH
17.
R. Rojas, Therorie der Neuronalen Netze, Springer-Lehrbuch, 1993.Crossref
18.
J. R. Koza, Genetic Programming, MIT Press, 1992.MATH
19.
R. Kruse, C. Borgelt, F. Klawonn et al., Computational Intelligence. Eine methodische Einführung in Künstliche Neuronale Netze, Evolutionäre Algorithmen, Fuzzy-Systeme und Bayes-Netze, Wiesbaden: Vieweg + Teubner Verlag, 2011.MATH
20.
Lomu.net, “50 Jahre Künstliche-Intelligenz-Forschung,” 2016. [Online]. Available: http://www.lomu.net/lomu7_KI-ueberblick.html. [Accessed 28 07 2016].
21.
P. McCorduck, Machines who Think: A Personal Inquiry Into the History and Prospects of Artificial Intelligence, San Francisco: W. H. Freeman, 1979.
22.
P. Koenig, “Künstliche Intelligenz,” 2007. [Online]. Available: http://www.heise.de/-280227. [Accessed 09 08 2016].
23.
A. Turing, Computing Machinery and Intelligence, 236 Hrsg., Bd. 59, 1950, pp. 433–460.
24.
P. Ford, “Unsere letzte Erfindung?,” 2016. [Online]. Available: http://www.heise.de/tr/artikel/Unsere-letzte-Erfindung-3152050.html. [Accessed 09 08 2016].
25.
Wikipedia, “RoboCup,” 2016. [Online]. Available: https://en.wikipedia.org/wiki/RoboCup. [Accessed 08 01 2016].
26.
E. Chin, “The Future of Fraud Fighting,” Sift science, 2015. [Online]. Available: http://start.siftscience.com/the-future-of-fraud-fighting-ebook?_ga=1.55925928.267213647.1472580662. [Accessed 08 01 2016].
27.
S. Bahrampour, N. Ramakrishnan, L. Schott and M. Shah, Comparative Study of Deep Learning Software Frameworks, 2015.
28.
Microsoft, “Internet der Dinge,” Azure IoT Suite, 2016. [Online]. Available: https://www.microsoft.com/de-de/server-cloud/customer-stories/thyssen-krupp-elevator.aspx. [Accessed 24 08 2016].
29.
S. Fernández Arregui, S. Jiménez Celorrio and T. de la Rosa Turbides, “Improving Automated Planning with Machine Learning,” in Handbook of Research on Machine Learning Applications and Trends: Algorithms, Methods, and Techniques, New York, IGI Global, 2010.
30.
P. Garica Laencina, J. Morales-Sanches, R. Verdu-Monedero and et al., “Classification with Incomplete Data,” in Handbook of Research on Machine Learning Applications and Trends: Algorithms, Methods, and Techniques, New York, IGI Global, 2010.
31.
Herschel, R.T. and Jones, N.E., “Knowledge management and business intelligence: the importance of integration”, Journal of Knowledge Management, Vol. 9 No. 4, pp. 45‐55, 2005.
© Springer-Verlag GmbH Germany 2018
Claudia Linnhoff-Popien, Ralf Schneider and Michael Zaddach (eds.)Digital Marketplaces Unleashedhttps://doi.org/10.1007/978-3-662-49275-8_31
31. How Banks Can Better Serve Their Customers Through Artificial Techniques
Armando Vieira1 and Attul Sehgal1
(1)RedOctopus Innovation, London, UK
Armando Vieira
(Corresponding author)
Email: [email protected]
Attul Sehgal
Email: [email protected]
31.1 The Impact of Big Data in Banks
Data generated from online bank transactions, digital sensors and mobile devices is being produced and recorded at an astonish rate. Every day 2.5 quintillion bytes of data are created and it is predicted that about 90% of all data processed was produced in the past two years [1].
The amount of data needed to be stored in servers is ever expanding, and for banks the opportunity to have a 360º perspective of the customer life has never been so high. These developments provide a huge business opportunity for banks, not only by improving their core business – risk assessment and investment – but also by creating new marketing opportunities and reducing costs. The challenge is how to extract the intelligence effectively [2].
Traditionally, banks have tried to extract information from a sample of its internal data and produced periodic reports to improve decision making. Nowadays, with the availability of vast amounts of structured and unstructured data from both internal and external sources, there is increased pressure and focus on obtaining an enterprise view of the customer in real‐time, on a self‐service basis and in a more systematic way.
By integrating predictive analytics with automatic decision making, a bank can better understand the preferences of its customers, identify customers with high potential to spend, be able to promote the right products to the right customers, improve customer experience, and drive revenue. However there are several challenges: i) engineering (how to store, organize and create views over the unstructured data in a cost‐effective way); ii) analytics (how to process real‐time analytics within an ever changing environment); and business (how to apply these insights to transform the business processes and translate into competitive advantages).
Advanced data storage for structured and unstructured data based on technologies like Hadoop and Spark – in the form of the so called “data lakes” – is becoming a standard. On the analytics side, Deep Learning based machine learning is making a tremendous progresses in extracting powerful insights from vast quantities of structured (transactional data), semi‐structured (social networks activity) and mostly unstructured data (images, video or text) [3–5] (see Fig. 31.1).
Fig. 31.1Deep Learning algorithms have a higher learning capacity
With online banking, credit card and mobile payment systems, banks have access to a large amount of customer information. Furthermore, social media data can shed light on brand sentiment and brand loyalty. However, most traditional customer behavior analytics techniques only focus on hard information and disregard soft and unstructured information.
Traditional customer behavior analytics have four dimensions: customer identification, customer attraction, customer retention, and customer development. Among them, customer identification is the most fundamental and widely implemented in the banking industry. Customer identification includes customer segmentation and targeting.
Each component of customer behavior analytics is linked to some standard data mining techniques. In customer identification, classification and clustering methods are usually used to target a specific customer group based on the business objective. In addition, regression techniques are applied to predict new potential customers. Almost all frequently used data mining techniques can be applied to better understand customer loyalty, including classification, clustering, sequence discovery, association, and regression.
Association techniques are often used in customer development in affinity analysis to find the relationship between different products that are bought by a given customer over his or her lifetime. Numerous solutions have been developed and studies done on traditional customer behavior analytics. For example, a framework for analyzing customer value and segmenting customers based on their value was proposed in [6].
31.2 Addressing the Opportunity
With the deriving of new customer insight through advanced analytics marketeers need to organize themselves internally to take full advantage of the available information. Here are some suggestions: 1. Build a Robust 360 degree Customer View – banks need to think beyond ‘one‐size‐fits‐all’ as relationship pricing and product bundling become ever more important. Banks should look at products and pricing based on CLTV – the value that customers bring to the bank across the spectrum of rates, fees, features and services. Despite the challenges in integrating data from multiple systems and data sharing impediments due to opt‐out policy and regulatory requirements, it is crucial to have a 360 degree view of customer to improve customer satisfaction and maximize lifetime value.
2. Adopt Advanced Customer segmentation – this is key to cater to individualized needs and should be based on standard banking metrics – tenure with the bank, number of accounts, balances of accounts and loans, frequency of interaction, behavioral (usage rate, price sensitivity, brand loyalty) and demographic variables (occupation, income, and family‐status). Traditional segmentation is dead ‐each customer is a segment.
3. Formulate Intelligent Real time cross–selling/up–selling Campaigns – banks can use real‐time events and customer insight to offer cross‐channel marketing campaigns where relevant events are acted on as a way to deepen customer relationships.
4. Design Innovative Reward Models – banks need to move beyond a “one‐size‐fits‐all” reward model and design a system where valuable customers enjoy premium benefits and redeem rewards points easily and in various ways.
5. Enable Automated Customer care Systems – In the digital era, customers demand more self‐service options, any‐time and anywhere. So expanding customer self‐service, case management, dispute management and event‐based decision‐making can be perceived as better customer care, while lowering operational costs and increasing effectiveness.
6. Enable Prediction through Big Data – Big data is the new disruptive technology. Big data technologies provides banks the ability to understand their clients at a more granular level and more quickly deliver targeted personalized offers. Being able to anticipate customer needs and resolve them before they become problems, allows banks to deliver timely, concise and actionable insight to contact center agents. This can also lead to increased sales, improved customer satisfaction and a reduction in operating costs. Fighting fraud, financial crimes and security breaches, in all forms, is among the most costly challenges facing the finance industry. Big data benefits include: Reduced costs of fraud screening and monitoring fewer false positives, reduced cost of fraud investigations, reduced payment fraud losses, real time fraud detection and mitigation, optimized offers and cross‐sell.
7. Improve Multi‐Channel Experience – banks should seek to attract and retain customers with a compelling multi‐channel experience across all touch‐points (branches, online, mortgage and investment advisors, etc.) Virtual channels are becoming more relevant, with the increasing penetration of high‐speed Internet connectivity and Web‐enabled mobile devices allowing consumers to spend more time online. Online banking and call centers already account for 55% of transactions [7].
Compared with traditional customer behavior analytics methods most used in banking, there are two major challenges in the big data era. The first involves how to handle the massive amount of complex data in a cost‐effective and efficient way. The availability of data has grown in magnitude, speed, and dimensionality. Second challenge is in new data analytical models required to capture the value behind the increasing amount of unstructured, soft information [8].
Traditionally, solutions to manage the large amount of data are unable to provide reasonable response times in handling expanding data volumes, leaving few options‐either to run the analytics models i
n batch mode or perform piecemeal transactions for a more reasonable response time. Therefore, a bank needs to ensure the real‐time response which requires new expertise in the data management and the latest systems management methods [9].
31.3 What Is Artificial Intelligence?
Asking whether a computer can think is a bit like asking whether submarines can swim (Edger Dijkstra).
In its simplest form, Artificial Intelligence (AI), consists of a set of algorithms that can perform complex cognitive tasks, some deem – up to now – being exclusive to humans, and makes them amenable to machines. AI is today one of the most exciting research fields with plenty of practical applications, from autonomous vehicles to drug discovery, robotics, language translation and games [5]. Challenges that seemed insurmountable just a decade ago have been solved and are now present in products and ubiquitous applications, like voice recognition, navigation systems, facial emotion detection and even in art creation – like music and painting.
Digital Marketplaces Unleashed Page 44
