Abstract
The precision marketing of commercial bank is among the most data driving and progressively confronting the difficulties of big data where phenomenal amounts of organized and unstructured data put away in an assortment of frameworks and arrangements — Identified as the best depicted utilizing volumes, variety, and velocity. Big data refers to the environment in which data sets to have grown too large to handle, managed, stored, and retrieved in a reasonable time.
The regulatory environment that commercial bank work inside requires these organizations to store and examine numerous long stretches of exchange data. Because of new guidelines, to acquire data which will influence the market entirely, precision marketing of commercial banks must concentrate on driving traffic towards their data frameworks.
This research paper was planned and completed to discover the precision marketing of commercial banks based on big data. Big data the related part of precision marketing of commercial banks procedure was investigated utilizing different extension that speaks to the various indications of banking vulnerability model. Considering the data mining model that can sufficiently handle the precision marketing of commercial banks dependent on big data utilizing the strategy will remove relevant client data from enormous accessible datasets and empower advertisers to create client division for doing direct or precision marketing for related items and administrations.
To achieve data mining purpose for this paper, it describes the commercial bank marketing data, data mining models, and feature selection. It equally attempts to find out how big data in commercial banks affect customer retention, detection, and prevention of fraud in the banking sector. This research additionally demonstrates the different systems used to model big data and big data that can be investigated utilizing different scientific techniques, for example, prescriptive analysis, descriptive analysis, and Predictive Analysis. This paper embraced the utilization of Regression models, which are the quality of predictive investigation. The connection between the dependent and independent variables tends to dissected the straight regression.
- Aims and Objectives
The main reason for this writing is to identify the effect of big data on precision marketing of commercial banks. Specifically, the issue here is to find the data mining model that can adequately handle the precision marketing of commercial banks based on big data. This technique will be the one that could extract meaningful customer data from large available datasets and enable marketers to develop customer segmentation for carrying out direct or precision marketing for relevant products and services.
The primary aim is to find the data mining model that will be employed to showcase how customer retention can be effective in the banking sector. It can equally be used to show how data mining could bring about the detection and prevention of frauds in the commercial banking sector.
The secondary aim is to find the data mining model that could effectively use supervised or directed learning algorithm for precision marketing in commercial banks based on big data. Hence, the process of learning is controlled by a formerly familiar dependent variable.
The third aim of this paper is to use the model that could bring about credit card approval of commercial banks customers. This will help to analyse the different scope of commercial banks process that represents the various signs of banking vulnerability model. The predictive model includes a method of how the software analyses some fact whereby the dependent variable is identified beforehand.
In a nutshell, an attempt will be made to explain the statistics, Logistic regression model, and another statistical formula that will be employed in the process of this research.
- Scope and Constraints
This paper is centred on the precision marketing of commercial banks based on big data.
It will specifically concentrate on how data mining model positively affects the precision marketing in financial sectors. The financial institutions have shifted from the traditional banking system to IT-based banking system where big data and precision marketing co-exist. This has significantly impacted on how financial sectors carry out their business operation and services.
The scope of this writing is to find the information mining model that can adequately handle the precision marketing of commercial banks based on big data. This technique will be the one that could extract meaningful customer data from large available datasets and enable marketers to develop customer segmentation for carrying out direct or precision marketing for relevant products and services.
From the available literature review, banks mainly focus on data mining methods to extract meaningful customer data from big datasets. Data mining model shows how customer retention can be useful in the banking sector and provide detection and prevention of frauds in the commercial banking sector.
The conceivable constraints in this research paper, precision marketing of commercial banks based on big data is that statistical formulae and tests lack reliability. Many authors have divergent views over the best model to be used in analysing the impact of big data on precision marketing on commercial banks. Another constraint to this study is that many models become obsolete in a short period.
- Introduction
The period of broad data is here, which is recognised as a disrupting technological difference in continuation to the cloud computing, including Electronic Power Control (EPC) system network. It has a significant impact on the community, human and businesses. Base on this writing, the precision marketing of commercial banks on big data will be analysed. Big data is the latest technological revolution in the Industrial Technology (IT) in the train of cloud computing and EPC system network (Santoso, 1996). Since the commercial banks’ development are involved, electronic trading has become the features in a period of broad data for financial industry growth. Because of this, commercial banks have to be positive in dealing with the current situation to face the challenges of the period of broad data from essential levels. A business organisation that promotes and create a data-driven improvement model will achieve higher productivity than others. These commercial banks can develop abilities of cross-sales and investment management efficiently, and at the same time, improve their core competitiveness of data. (Moin, 2012).
Big data was a major concern years before 2008 in the business community of China, the academia and it has been a factor which drives certain countries such as China to innovate more and compete better with other countries. Since the discovery of big data, it had proved itself to be an essential tool which can be used to develop the community both socially and economically. However, it should be noted that precision marketing can only be achieved through the help of big data. Precision marketing which focuses on establishing interactions between marketers and consumers. With precision, marketers have complete access to their consumers to analyse and predict data
Big data has its actual worth as the substantial potential commercial worth. (Zhao, 2009) Stated that the western financial and technical data shows that productive use of broad data could develop margins for most significant retailers, commercial banks and also human, society and businesses up to 60%. Therefore, how to take advantage of these big data to propel administration determinations are displaying the focus of most financial institutions and companies. In recent times, the purpose of intelligence technology concentrates mainly on consumer relationship administration, dynamic pricing and market segmentation, among others. It has established that an expert marketing system based on the advanced intelligent network has described some differences in creative marketing ideas and patterns (Ping, 2013). In line with this, (Bughin, 2010) asserted that a cross-platform smart marketing guide based on open-end funds; expand the fund’s operation spans in the side and upward areas, analysing to combine with the discussion delivery, fund marketing and after-sales.
Unlike the manufacturing and industrial sectors, the commercial banks have no making or logistic methods. Instead, the daily operations of commercial banks, insurance companies, security firms and other financial institutions are serviced by a massive number of real-time business dealings, with each one backed up by the significant amount of big data to sustain decision making. The possible value of Big data is considerable and always bring changes to the financial industry (Demirguc-Kunt, 1998).
The changes will start appearing in two areas. In the first place, in precision marketing, where broad data change the information formation, it allows commercial banks to collect and analyse customer’s data to give more individualised and tailored assistance. Lastly, in risk management, big data alters the standard risk management standard with cloud computing, which can improve the most precise risk evaluation at the most moderate cost (Chitra, 2013). Commercial banks that master this data-driven economy could benefit from improved operational effectiveness and business performance. The business institution could continue to shift from universal management at the macro level to micro-management of comprehensive processes and performance.
Big data encourage the growth of internet banking. Its focus is the spread of data, which make the costs of business dealings and analysis decrease significantly. Currently, the entire business model has changed. The differences offer a new way for the commercial banks’ development to maintain excellent customer practice and save cost, which helps increasing consumer investment yields and lowering loan interest rates (Kaptan, 2002). Hence, the banks could draw more customers and develop consumer volume. In contrast, traditional banking pays more attention to a few customers, such as big-scale enterprises and higher-yielding customers.
- Literature Review
Traditionally commercial banks conducted their marketing activities through mass media channels such as print media, radio, television, etc. (O’guinn, 2008), however, studies have revealed that these techniques are ineffective in today’s world, as they do not develop a direct relationship with the customers who are increasingly uninterested in traditional types of sales promotions Petrison et al. (1997). Therefore, banks and insurance companies have shifted their focus towards direct or precision marketing strategies for more effective interaction with customers Ou et al. (2003). Various channels have been identified by researchers for carrying out precision marketing such as Email, telemarketing, personal contact, etc. (Elsalamony, 2014). However, these marketing channels are dependent upon big data analytics and algorithms for carrying out precision marketing.
Wisaeng (2013) has analysed that banks mainly focus on data mining methods to extract meaningful customer data from large available datasets. The data mining techniques enable marketers to develop customer segmentation or profiles for carrying out direct or precision marketing for relevant products and services. Moin and Ahmed (2012) have identified various data mining techniques, such as class, association, clustering, prediction, and sequential patterns. Ramager (2010) have discussed that classification technique is most effective for modelling customer records in pre-defined segments. Similarly, Elsalamony (2014) has analysed various algorithms or models for carrying out classification data mining such as Multilayer Perception Neural Network (MLPNN), Bayesian classification, Decision Tree Model, and Support Vector Machines. This technique can help marketers segment their customers based on different products and services.
Bhambri (2011)has studied association or correlation data mining technique, which is most important for extracting data for marketing purposes, as this technique enables marketers to detect patterns in data. The models logically connect different events and enable decision-makers to develop effective marketing strategies to target different customer segments in a more personalised way. Kaptan and Chobey (2002)have studied that association technique can enable marketers to optimise cross-marketing strategies by carrying out effective analysis of customers shopping behaviour. Ramager (2010) has identified various models for association data mining such as multi-level association, multidimensional organisation, quantitative organisation, direct organisation, and indirect organisation.
Clustering is another vital data mining technique for decision marketing. Raju et al. (2014) have analysed that it allows marketers to group customers in clusters or classes having similar buying behaviour, set of queries, services requirements, transactions, etc. Karguptaet. Al (2005) have discussed that clustering employs a pre-processing approach to classify customers depending upon their geographical location, the demand for services, policy awareness, etc. In this way, banking sector marketers can segment customers in clusters depending upon the data instead of classifying them in pre-defined segments. This technique enables banking marketers to carry out class-oriented marketing instead of mass marketing to all customers.
Similarly, it also effectively facilitates cross-marketing. Chitra and Subashini (2013) have discussed the prediction or forecasting technique for future precision marketing in the banking sector. This technique will enable marketers to learn from current data and using regression models such as logistic regression, neural nets, decision trees, etc. forecast customer behaviour in different segments. In this way, commercial banks can stay ahead of the competition through effective learning from the data and precise marketing to customers evolving demands.
Various studies have analysed the potential of big data precision marketing for commercial banks. Muraleedharan (2009) has discussed that big data enables marketers to recognise factors that influence customer requirement for different products and services and evolve their future marketing strategies in a more precise manner. In this way, marketers can adopt a customer-oriented approach and techniques for efficiently promoting their products. Bhasin (2006)has analysed that big data can help commercial banks in determining the segments of customers which might be more interested in future product offerings and investing marketing budget towards these segments in a more specific and precise manner. This will enhance the effectiveness of the marketing campaign in terms of achieving higher conversion and customer retention rate.
Petry and Zhao (2009) have analysed the cross-selling advantage of precision marketing, which enables banking marketers to learn from customer data and recommend them products similar to the ones that they have previously purchased. In this way, they can maximise the profits from the existing customers and especially from the most valuable customers who are the source of major revenue for the bank. Dass (2011) has discussed that big data precision marketing can significantly improve the response rate of the customers since through advanced data mining techniques; marketers will be able to identify the relevant consumer sectors for various products and services and operationalise the marketing campaigns accordingly. In this way, the relevant customers will be reached in a short time, and their response or lack of response can be recorded as well. This will increase the efficiency of the sales force and enable banks to maximise their revenues.
Hu and Liao (2011) have discussed that big data can enhance the efficiency of electronic and telemarketing. Moro et al. (2014) have analysed the success of data-driven bank telemarketing by conducting a study of Portuguese banks and found out that an intelligent Decision Support System (DSS) based on data-driven models for carrying out precision telemarketing is necessary for the success of the campaign. Four data mining models, i.e. logistic regression, decision trees, neural networks, support vector machines (SVM) were analysed, and it was found that the success of marketing campaigns can be enhanced by 79% if the banks contacted only 50% relevant customers through data sampling instead of reaching all the clients.
A more careful review of the literature reveals that studies mainly focus on data mining techniques and the potential of big data in marketing while there is a lack of dedicated studies analysing the viability of precision marketing by evaluating real-world data of the commercial banks. This gap can be filled by analysing commercial banks employing big data precision marketing techniques through different channels and assess their success against different data mining models.
- Data and Methods
This paper will attempt to analyse the impact of precision marketing of commercial banks based on big data. This section will describe the commercial bank marketing data, data mining models and feature selection.
However, the method that will be adopted for this paper will be data mining models. It will equally attempt to find out how big data in commercial banks affect customer retention, detection and prevention of fraud in the banking sector.
3.1 Commercial Bank Marketing Data
We all hear of how big data is making changes, especially in commercial bank marketing systems. Most commercial banks are saddled with the problem of where to get their data and how they can effectively leverage this data. Demographic data which gives information about customers’ age, income, marital status, gender etc. are often gotten from the customer resource management sector of commercial banks and are commonly used by these banks. Many commercial banks do not have a central source of data which can be efficiently utilised. Hence, models such as Google Adwords can be easily used. Google Adwords data will show you topics that are trending whether it is “tax reform”, different types of accounts such as savings account etc. To obtain data which will affect the market positively, commercial banks must focus on driving traffic towards their data systems.
One of the tactics used in Florida is the purchase of common search terms by people. Banks such as community banks buy these commonly search terms, use it. Next to the name of their company or probably run ads which will be viewed by all. On the long run, what this ad does is getting the bank companies new customers who are willing to give the banks a trial. Specific terms may provide the most reliable impressions but will not get you most clicks if you use generic terms. For this reason, banks must make use of some target terms for effective use of marketing data.
The data size of commercial banks after many years’ growth and development has increased drastically, while the amount of unorganised data is multiplying at the same time. The application of big data has changed the function of financial agents. This leads the commercial banks to achieve higher productivity because of the improvement in marketing and digesting data (Zhang, 2016)
The virtualised transactions of financial products and services could upgrade the commercial supply chain since the financial expenses can be decreased while the economic market efficiency can equally be improved. This means that the accumulation of marketing data makes the commercial bank’s sales more precision. Based on existing customers, commercial banks can discover more valuable clients who have the prospective to market accurately. Using commercial banks, for instance, big data can help them to ascertain an active and dependable credit system to recognise high-risk customers and prospective customers. This leads to the identification of various trading risks that can prevent and control the financial risks of customers. Hence, in the era of marketing data, commercial banks innovation will be rapid. By using big marketing data, commercial banks can improve their profitability (Groll, 2015).
3.2 Data Mining Models
When creating a data mining model, it is usually advisable to use an algorithm to data. A good data mining model enables its users to make reasonable predictions, generate inferences about relationships making use of a set of data or statistics. The different mining techniques involves tracking processes, classification, association, detection of outliers, clustering, regression and prediction.
Although, data mining models may appear similar to standard data tables, but they differ from them in that mining models are virtual images. Structurally, a Data mining model should have a data mining column together with the data mining algorithm to enhance predictive analysis.
When training a mining model, it is vital to feed the information that was existing with recent trends and store the results rather than raw data. To build a mining model, be sure you are embarking on a tasking journey.
To build a mining model means the body must first recognise and decide on what they want their model to look like, plan for the processes involved in its successful building to give an accurate prediction analysis. Mining models that become irrelevant are those models that are not practical or those who are too costly to implement. Together with other factors such as deployment environment, banks must strive to design models which will be useful, identify missing analysis tools and produce the best result needed for accurate predictive analysis.
Data mining principles are a central part of the idea of information mining and are of virtual structures to portray data grouped for predictive investigation. Data mining principles might seem to be very related to data tables, but tables are used to illustrate actual findings of data, whereas mining guides are interpretations of those data or cases. This information store factual information that depicts the guidelines and examples gained from preparing the model. The model is developed by bolstering existing data and patterns to it. Cases are assembled to frame together case sets, which make up a mining model (Spotfire Blogging Team, 2013).
An information mining model is fundamentally made out of a few information mining sections and an information mining calculation. The substance created at the point when the model was prepared put away as information mining standard hubs. It is necessary to perceive that the information used to develop the model are not put away with it. It is just the outcomes that are put away. This structure empowers enormous volumes of data to be utilised during the preparation technique; in this manner, it helps in improving the conceivable effectiveness of any expectations made by using the model.
3.3 Feature Selection
Selecting a feature for a subset is an critical problem which must be solved in discovering knowledge. When choosing features for analysis, it is important to eliminate variables with little or no information for prediction. This problem is a key factor in improving the understanding of models built for data analysis. For accurate predictive analysis, it is important to select features which will.contain variables that best explain models and give insights into a better understanding of analysis.
Feature selection comes in various forms starting from those with supervised learning to others with unsupervised learning. The supervised learning is used to give a higher level of classification accuracy. In unsupervised learning, feature selection involves the use of learning algorithms to find natural groups of data. In this line, what feature selection does in unsupervised learning is to find subsets of features comprising of functional clusters which are of the high-quality form a number of given clusters.
The use of a single criterion for feature selection is a traditional way which is not completely accurate as users are supposed to use multiple criteria in the evaluation of their feature selection and decision-making process. To achieve a clearer picture of result amongst various objectives available, feature selection has been formulated to be a multi-objective optimisation.
In supervised learning, direct mailing of potential customers has proved to be a good way in sourcing data from various customers and marketing a new product or service as well. When companies make use of this model, they tend to reduce cost and expenses as target customers are being reached rather than wasting money on random individuals.
In unsupervised learning, feature selection involves clustering, which will improve customer relationship management. This model can not provide new models for marketing because the model makes use of subset features based on a piece of previous knowledge about the market.
Featured collection is the method whereby one automatically or manually selects the features that contribute most to the expectation variable or yield in which one is interested in. Having irrelevant characteristics in your data can reduce the efficiency of the models and delegate your model to learn based on unnecessary features.
- Experiments and Results
4.1 Modeling
There are various techniques which can be used to model big data. Some methods involve the use of a big table, such as those used by Google to store comments on social media. The big table is referred to as a key table having a certain number of rows designated as n. Each row usually has its key identifiers, and there are columns which are also important features of the model.
Modelling explains the design of a system logically and explains how a database can be implemented effectively. Over 80% of big data is unstructured and not organised; hence, there is a need for data modelling to play an important role in data analysis.
This paper will make use of Regression models, which are the strength of predictive analytics. The relationship between the dependent and independent variables tends to analyse the linear regression. Based on the analysis, big data can be analysed using various analytical methods such as those described below;
Prescriptive analysis
This type of analytics is usually used to determine the cause and effect relationship among business optimisation policies and analytic results. To maximise business models, prescriptive analytics make use of information gotten from predictive analytics. Several organisations do not use this model because most databases have a constraint on the number of dimensions they can capture (Banerjee et al., 2013).
Prescriptive analytics is tied in with utilising data and analytics to improve decision and consequently the viability of activities. Prescriptive analytics is a rising order and represent a further developed utilization of predictive analytics. Prescriptive analytics goes past mostly anticipating option in the predictive model and proposes a scope of recommended activities and the potential results of each activity
Descriptive analysis
Descriptive Analysis is brief expressive coefficients that abridge a given data collection, which can be either a representation of the whole or an example of a populace. Descriptive Analysis is separated into measures of central tendency and measures of variability (spread). Measures of central tendency incorporate the mean, median, and mode, while measures of variability include the standard deviation, variance, the minimum and maximum variables.
These type of analysis are known to be the simplest form of big data analysis because they make use of statistical methods such as mean, median, mode, variance, standard deviation etc. to describe and evaluate knowledge patterns. Nowadays, descriptive data make use of information from predictive analytics.
The most perceived kinds of descriptive analysis are measures of focus: the mean, median, and mode, which are utilised at practically all degrees of math and insights. The mean, or the average, is determined by including every one of the figures inside the data set and then isolating by the quantity of values inside the set. For instance, the whole of the accompanying data set is 40: (4, 6, 8, 10, 12). The mean is 8 (40/5). The mode of a data set is worth showing up frequently, and the median is the figure arranged in the data set. It is the figure isolating the higher numbers from the lower numbers inside a data set. Notwithstanding, there are less-normal sorts of descriptive analysis that are still important.
Individuals utilise descriptive analysis to repurpose difficult to-understand quantitative bits of knowledge over a vast data set into reduced down portrayals. A learner grade point average (GPA), for instance, gives a decent understanding of descriptive analysis. The chance of a GPA is that it takes data focuses from a broad scope of tests, classes, and grades, and averages them together to give a general understanding of a student’s general scholarly capacities. An understudy’s close to home GPA mirrors his mean scholastic performance.
All descriptive analysis is either a measure of central tendency or measures of variability, otherwise called measures of scattering. Measures of central tendency spotlight on the average or centre estimations of data sets; though, measures of variability centre around the dispersion of data. These two measures use diagrams, tables, and general dialogues to help individuals understand the meaning of the dissected data.
Predictive analysis
This model is used to determine future possibilities making use of supervised, partially supervised or non-supervised learning models. This model is based on statistical models and focus on revealing relationships between data obtained. Predictive analysis is grouped into two groups, which are the regression techniques and machine learning techniques. Linear regression makes use of the correlation between all types of variables to make predictions. The model explained in this book is the logistic regression analysis.
It’s critical to perceive that this analytics is about probabilities, not absolutes. When applying Predictive analysis, one doesn’t know ahead of time what data is significant. Predictive analysis figure out what information is predictive of the result you wish to foresee
4.2 Logistic Regression Analysis
With logistic regression analysis, the correlation between a dependent variable and a set of independent variables can be studied. Logistic regression is a type of regression analysis used for predicting the result of a particular dependent variable based upon one or more independent variables. Instead of providing the data to a straight line, logistic regression uses a logistic curve. The method is as shown below.
(1). The logarithmic formula can be used to get the logistic function, as shown below.
(2). Logistic regression provides excellent performance on a broad type of problems. Akaike Information Criterion (AIC) is given as follows.
AIC = -2log L+ 2p
(3). Where: L is the maximum likelihood of these fitted model and p is the number of estimated parameters.
In a decisive step and after taking the model with the weakest AIC value, we calculate the test chi-square by the logistic model selected. The amount of chi-square is provided by:
4.3 Predictive knowledge and potential impact (Result)
In this paper, eighteen (18) ratios correlated with various dimensions of commercial bank analysis that represents the multiple indicators of banking vulnerability measures are used. The proportions are regrouped into five (5) groups such as liquidity, management, activity, profitability and vulnerability as shown in Table I. Table II shows the descriptive statistics, while Table III shows the selected variables according to the criterion of the lowest AIC information. Table IV shows the results of logistic regression and Table V shows the Odds Ratios from the statistical data.
4.4 Discussion
The utilisation of the well-ordered procedure in the decision of the most separating factors indicates a foundation of the weakest descriptive with a measurement of 71.049 for the model that regroups the elements X5 to X14 which are introduced in Table II and III.
In Table IV, every one of the factors is measurably critical to the edge of 5% aside from the proportions X9 and X13 that are important at 1 %. Then again, the model is all-inclusive substantial to 1 % with a measurement of 34.090.
The X5, X8, X9 and X12 ratios are contrarily significant, and the estimation uncovers that the variable that estimates the capacity of commercial banks to repay their debts (X5) reduces in an important way for the likelihood of deformity for the commercial banks. In the same vein, the variable X8, X9 and X12 that are the commercial banking exploitation coefficient respectively, the deficient banking likelihood, the financial gainfulness and the proportion of budgetary move, contrarily impact the inadequate likelihood.
In the same way, the ratios X6, X13 and X14 show the misuse costs, the productivity and the capacity of the commercial bank to repay its obligations, spare the positive centrality and after that expansion the likelihood of the bank deficiency respectively.
The odds ratios study, as shown in Table V, reveal that the commercial bank profitability X13 records high ratio that is greater than 1. This means that the deficient response is strongly probable.
However, the probability of deficient response is strong for the ratio of exploitation costs X6. The increase of banking profitability and then a balance of assets liabilities puts the banks in an excellent performance.
The proportion X14 that estimates the capacity of a bank to reimburse its obligations records a significantly high proportion, and the proficient reaction is additionally high. It is apparent that, if the size of the commercial bank savings and banking capital increases, the bank is then found able to honour its commitment towards their lenders. Hence, the capacity of reimbursement drives banks into excellent performance towards their customers.
- Conclusion
The use of a data mining model and logistic regression has shown that the adequacy of precision marketing in commercial banks depends on big data, the analytical variable in the logistic model has produced enough results with anticipated signs and significance. Furthermore, the most appropriate proportions in the description of banking unwavering quality at the Commercial banks are the expanding of banking products and the strength of banks to refund their money which seems to be a high odd ratio.
Mining models are used in a wide range of applications including political forecast and weather pattern detecting and to rank page webs amongst various other uses.
The effective utilization of big data can change economies, conveying another influx of profitability development and consumer surplus. It is likewise expected that throughout the following five years, there will be striking and critical advancement dependent on big for data precision marketing of commercial banks. Commercial banks are starting to work out guides of where big data could convey the most incentive inside this more extensive arrangement of accuracy advertising.
Utilizing big data will turn into a key of rivalry for existing organizations, and will make new contender who can pull in workers that have basic abilities for a big data world.
- Appendices
7.1 Figures
Fig. 1: A chart showing data mining techniques
| Data Mining Techniques |
| Problem Understanding |
| Data Understanding |
| Data Filtering |
| System Modeling |
| System Evaluation |
| Analyzing Results |
| 1. Knowledge about commercial bank |
|
2. Its challenges |
| 1. Gathering of data about the needed attributes |
|
2. The client data |
| 1. Preparation of the attributes for data filtering |
| 2. Selecting, checking and making data ready to be mind |
| 1. Customer retention, fraud detection and prevention for system |
| 2. Checking applicable attribute |
| 1. New customer data checker with designed model |
|
2. Result verification |
|
|
| 1. Comparing Models |
|
2. Performance analysis |
Fig 2: A chart showing data mining processes in commercial banks
| Data Mining processes in Commercial banks |
| Customer Retention |
| Fraud Prevention |
| Fraud Detection |
| Supervised Learned |
| Credit Card Approval |
| Clustering |
| Decision Tree |
| CART |
| Logistic Regression |
| SVM |
| Decision Tree |
| EM Algorithm |
7.2 Tables
Table I: The financial ratios according to big data
| Variables | Financial Ratios | Category |
| X1 | Liquidity | |
| X2 | Liquidity | |
| X3 | Liquidity | |
| X4 | Liquidity | |
| X5 | Liquidity | |
| X6 | Management | |
| X7 | Management | |
| X8 | Management | |
| X9 | Management | |
| X10 | Activity | |
| X11 | Activity | |
| X12 | Activity | |
| X13 | Profitability | |
| X14 | Vulnerability | |
| X15 | Vulnerability | |
| X16 | Vulnerability | |
| X17 | Vulnerability | |
| X18 | Vulnerability |
Table II: Descriptive Statistics
| Variables | Min. | Median | Max. | kurtosis | Stand. Dev |
| X1 | 0.000 | 0.794 | 6.248 | 80.511 | 0.040 |
| X2 | 0.000 | 0.872 | 1.010 | 5.068 | 0.021 |
| X3 | 0.000 | 1.061 | 12.081 | 43.591 | 0.101 |
| X4 | 0.000 | 12.24 | 6146.411 | 104.991 | 43.942 |
| X5 | 0.000 | 0.052 | 1.713 | 71.049 | 0.011 |
| X6 | -0.078 | 0.032 | 0.279 | 59.469 | 0.000 |
| X7 | 0.000 | 0.071 | 0.651 | 97.531 | 0.001 |
| X8 | -0.488 | 0.401 | 0.989 | 7.800 | 0.022 |
| X9 | 0.000 | 72.81 | 287.582 | 5.021 | 3.169 |
| X10 | 0.000 | 0.832 | 8.433 | 115.512 | 0.050 |
| X11 | 0.000 | 0.922 | 2.101 | 7.149 | 0.021 |
| X12 | -1.159 | 8.771 | 82.113 | 21.658 | 0.833 |
| X13 | 0.000 | 0.041 | 0.379 | 72.644 | 0.000 |
| X14 | 0.000 | 0.553 | 3.191 | 0.589 | 0.070 |
| X15 | 0.000 | 0.981 | 2.131 | 7.281 | 0.021 |
| X16 | 0.000 | 0.00 | 0.119 | 13.311 | 0.000 |
| X17 | -0.341 | 0.101 | 1.532 | 19.479 | 0.021 |
| X18 | 0.000 | 72.691 | 1504.089 | 19.481 | 18.289 |
Table III: Selected variables according to the criterion of the lowest AIC information
| Variables | Min. | Median |
| X5 | Liquidity | |
| X6 | Management | |
| X8 | Management | |
| X9 | Management | |
| X12 | Activity | |
| X13 | Profitability | |
| X14 | Vulnerability |
Table IV: Results of Logistic regression
| Coefficient | Erreur Std. | z | p- critique | |
| Constant | 0.141982 | 0.823611 | 0.1724 | 0.86313 |
| X5 | -52.5969 | 22.9047 | -2.2963 | 0.02166** |
| X6 | 167.726 | 85.2452 | 1.9676 | 0.04912 ** |
| X8 | -12.62 | 6.16244 | -2.0479 | 0.04057 ** |
| X9 | -0.0732672 | 0.0254852 | -2.8749 | 0.00404 *** |
| X12 | -0.297681 | 0.137488 | -2.1651 | 0.03038 ** |
| X13 | 114.552 | 44.4356 | 2.5779 | 0.00994 *** |
| X14 | 3.1964 | 1.50566 | 2.1229 | 0.03376 ** |
| = -27.52638 | ||||
| -2 = 55.05277 | Chi-square, = 34.08966 P-value = 0.00002 | |||
Table V: Odds Ratios
| Odds Ratios | |
| Intercept | 1.15256 |
| X5 | 0 |
| X6 | 6.956 * 1072 |
| X8 | 0 |
| X9 | 0.92935 |
| X12 | 0.74254 |
| X13 | 5.622 * 1049 |
| X14 | 24.44436 |
Students working on case studies or might need academic help, might find our custom Case Studies Writing Services helpful.
For more academic help please check a wide range of services our Finance Writing Help team offers:
– Finance Assignment Help
– Finance Essay Writing Services
– Finance Dissertation Help
– Buy A Finance Research Paper
– Financial Management Assignment
