Value Creation Using Big Data Mining In Banking

Introduction: Problem Statement

The banking sector deals with an overwhelming amount of data on a day to day basis. Banks have historical records and customer information data dating back to decades. Manually leveraging such enormous amounts of data for making decisions is not only time consuming but involves high risk and can be fallible. In addition to this, with the growing banking industry, there is an increase in the fraud rate as well. Therefore, the security aspects of banks must be dealt with on a priority basis. This is where big data techniques come into play. Big Data techniques such as Data Mining (DM) is crucial for the banking sector. It helps create value from the large amount of data in hand and thereby facilitates better decision making and helps achieve better strategic management and customer satisfaction.

Data Acquisition

The research data has been recorded over the years 2013-2018 and describes over 100 recent DM applications in the banking sector. A variety of techniques have been used by different banks. Security and fraud detection: To detect frauds, customer transactions are investigated to see if there is anything that indicates fraud or opportunities for fraud to be perpetrated.

Customers are categorized as good or bad depending on their repayment performance over a period. In addition to this data from loan applications, personal /business credit records etc of the customer are analysed to evaluate the potential risk of lending money to that customerCustomer Relationship Management: For customer profiling and segmentation, a survey of users was conducted, and this was used to analyse the attitudinal, behavioural and demographic variables of the customer. Report OverviewThis report aims at creating value out of the enormous amount of data that banks possess by making use of DM techniques. Data is the most asset, especially in financial industries. The value of this asset can be evaluated only if the organization can extract the valuable knowledge hidden in raw data. (Balan, 2013). The solution to this is Data Mining. It has been observed that banks adopt DM techniques for the following purpose:

• Security and fraud detection.
• Risk Management and Investment banking
• Customer Relationship Management

The report also gives a summary of the various statistical software’s used by banks to analyse customer behaviour, identify the potential customers etc.

Data Preparation

Security and fraud detection: Lingo clustering technique was used for the pattern matching process and a hybrid DM using K-means algorithm was used to predict network intrusions. To decide whether a fraud has taken place or not a semi supervised decision-making system (which makes use of machine learning) called Bank Sealer had been used. Risk Management and Investment banking: Clustered SVM Classifier and Multilayer Perceptron NN model that uses the back-propagation algorithm were used by banks for credit card scoring. To detect bad credit scores for peer to peer lending (P2P), Logistic Regression Analysis and K- means clustering technique were used. CRM: For proper segmentation of customers RFM analysis model was used. RFM measures the Recency i. e. how newly the customer has procured, the Frequency - how regularly the customer procures and Monetary - how much amount the customer pays out.

Findings: Data Analysis The data analysis method of our journal according to the writer has been based on the accounting framework of Data mining (DM) techniques mapping shown below. Hence, the writer further emphasised the use of keywords such as banking, fraud detection, credit card, credit scoring, risk management, deposit, mortgage, debit, loan, CRM, bank marketing; and data mining, clustering, text mining, classification amongst others DM methods to process and analysis the numeric and non-numeric data-. Nonetheless, to ensure completeness of the findings, the use of corresponding similarities search tactic was used so that every aspect in covered in the analysis. Results & Insights Sector References Key Techniques Regions Purposes Security and fraud detection [6,10–32] classification (DT, NN, SVM, NB), k-mean clustering, ARM Australia [12], Latin-America [29], Greece [24], Germany [32], Belgium [21], UCI Repository [15–18] Identifying phishing, fraud, money laundering, credit card fraud, security trend of mobile/online/traditional banking. Risk management and investment banking [33–51] classification (DT, NN, SVM, NB, LR), k-mean clustering UCI Repository International Dataset [41,42,45], Australia [51], Iran [37], Indonesia [34], China [35], German [36,38,39,51], Taiwan [51], US [49], Canada [46] Credit scoring, credit granting, risk management for peer-to-peer lending. Customer profiling and knowledge [52,53] classification (DT, NN), k-mean clustering Jamaica [52] Efficiently build accurate customer profiles Customer segmentation [54,55] k-mean clustering Iran [54] Provide sufficient customer segmentation, conduct customer-centric bus Customer satisfaction [56] k-mean clustering, classification (NN) Spain [56] Make the most strategic investment on maintaining and enhancing customer satisfaction Customer development and customization [57–68] classification (DT, NN, NB, LR, SVM), k-mean clustering Portugal [57–62], Turkey [66,67], China [68], Taiwan [65], UCI Repository [64] Strategic banking via direct marketing, targeted marketing, product cross/up selling Customer retention and acquisition [69–75] classification (DT, NN, LR, SVM), ARM, k-mean clustering EU [69], China [70], Nigeria [71], Croatia [73], Bangladesh [75] Customer churn prediction and prevention, attracting potential customers and strategic future service design. Others [76–84] classification (NN, DT, SVM), k-mean clustering Nigeria [77], Turkey [78,81], Canada [80], ASEAN [82], Islamic banks [83], BRICS [84], US [79 Branch strategy, bank efficiency evaluation, deposit pricing, early warning of failing bank.

Data Presentation

Exploratory data was conducted to identify DM techniques for identifying fraud behaviours and transaction patterns; using topics, DM techniques and various software’s.

After the review of 100 publications according to the authors of our reviewed journal. Our findings from the publicized journals revealed Big Data techniques of Data Mining (DM) techniques are integral and adopted in banking to analyses enormous data generated in banking for strategic and meaningful decision making such as cluster analysis, association rule mining (ARM) and classification techniques, which include but are not limited to Decision Trees (DT), Neural Networks (NN), Support Vector Machines (SVM) Naive Bayes (NB) and Logistic Regression (LR). Furthermore, our findings also unveil the manner of DM applications in banking scenario in line with fundamental data to ensure, Security and fraud detection, compliance through risk management and investment mentoring to ensure conformity with standards in banking, Despite, the process of extracting information manually there are restriction on the outcomes of the research finding in statistical and diagrammatic presentation of the various findings due to confidentiality restrictions and generate needful information, customer segmentations and product(services) development for appropriate customers segments.

Conclusion: Actions

Based on the review of over 100 journals, the authors applied different data mining techniques for managing risk, fraud detection and CRM in the banking sector. Thus, it has been successful despite exception on big data application because of different variable features. The author used different techniques and software to get an insight for development of data mining for banking in future, there was special emphasis on the obstacle and challenges on impact of big data.

Our perspectives

The study clearly analysed development of DM applications in banking by giving insight to future research and challenges by noting the unstructured big data could be cumbersome to explore and thereby showing the DM technique that could be exploited in tackling the issue. E. g. text mining, entity extraction and social network analysis through clustering and classification which clearly shows enough evidence of both application and its popularity. The study also shows that Weka, MATLAB, SPSS, R and RapidMiner seems to be the most popular software adopted by banks which is used for fraud detection, risk management and CRM although most application used more than one technique. DM applications in banking has helped in value creation through customer information gathering and these generate insight into the potential and valuable information yet undiscovered.

Concerns & Limitations

The study is basically focused on creating value and use of data mining technique and financial technology in the banking industry however our major concern is that the study seems to have ignored that the use of expensive software or data mining is not enough in data science and it will be better if the study focus more on the embedded analytics and statistical model which is very important. Machine learning, deep learning is all scientific approach, so attention should also be given to the statistical approach. This is the era of big data and research directions of the applications of DM in banking keeps changing because of the amount of data generated, the cost of purchase and maintenance of this new technology might affect the banks.

The study should also focus on privacy issues as it relates to the new GDPR guidelines which involves banks collecting information about their customers through text mining to understand the trends in their purchase attitude which is very susceptible to being stolen through hacking. Data mining has lots of benefit for the banking industry, attention should be given to security, privacy and information misuse and therefore we suggest that this should be handled with care and all banks must therefore be compliant with the new GDPR guideline and be more focused on internet security. Finally, the techniques used by the banks for data mining has provided the accuracy of its own limits a typical example is the Cambridge analytical getting information of people through text mining from the social media and using such information against the wishes of the people during the Brexit referendum.