Banks' Techniques of Data Mining Essay

Banks adopts different techniques of data mining to manipulate large quantity of data stored in the data warehouse to create models that create valuable assumption Those techniques include association whereby a pattern is discovered from a relationship between items in the same transaction. For example a bank manager discovers that a particular customer withdraws, deposit, transfers money at a specific branch of the bank. Therefore the manager will always associate the customer with that particular branch and incase a transaction is conducted in another branch, the system will detect and a lot of precaution measures will be used to ascertain whether it is the original owner of the debit card performing transaction.

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Save 25%

Banks also use classification technique, whereby it applies mathematical techniques such as linear programming, network diagrams and statistics to classify data items into predefined set of groups (Berson & Smith 2001). For example using an application software a prediction of transition of junior bank accounts to either current or fixed accounts can be made because age is a determinant variable. Banks can also predict outcomes of advertisement of its services, for example whether new customers will be enlisted in the bank. In addition to that, banks can also detect fraud through use of probability techniques of evaluating probability of a customer using another branch of the bank.

In addition banks use clustering technique which defines classes and puts objects in each class and thus end up making a useful cluster of objects with similar characteristics. For example banks detect fraud by using past methods that were used by criminals to withdraw money from customers account. Therefore bank's system detect whenever a duplicate debit card is used and also when a transaction is conducted in an unusual place and in unusual time.

Moreover, bank's fraud detection system uses prediction technique which applies both independent and dependent variables to discover relationship between them given one independent variable. For instance, prediction technique is used in anomaly detection while investigating cases of fraud on manipulated data whereby financial data is the independent variable. For example a bank manager can detect fraud if a debit a holder conducts transaction in different branches of the bank within a time interval which is not even enough to reach the location. This can indicate use of a duplicate debit card.

In addition, bank's fraud detection system uses sequential pattern to identify similar patterns on regular events during banking transactions, for example, a bank managers can predict customer's pattern of either withdrawal or deposit of cash. But if a customer at once make a request to withdraw all the cash, the bank manager can predict possibility of illegal transaction by a criminal.

Lastly, bank's fraud detection system can use decision tree which is a very common data mining technique that utilizes root node, branches, internal node and leaf node to derive different questions and answers that assists in making final decision of possibility of fraud occurence. The topmost is root node which has a condition with multiple answers. It is then followed by branches which have both answer of root node and a question to the internal node.

Banks use decision trees because they are simple; they do not require domain knowledge, they use non-linear data structure. In addition decision trees are easier to comprehend.

Data mining plays significant roles in different industries; for example in marketing, it assist retailers to predict purchase patterns of their customers and thus ensure there is enough inventory. It also assist government to detect cases of money laundering and siphoning basing its argument on the data that is fed in exchequer account and also the pattern of withdrawal. Data mining also assist banks to ensure security of their customers' accounts; by detecting pattern of withdrawal and location of withdrawal basing their arguments on location and frequency of withdrawal.

Data mining also have some weakness, for example privacy is a major issue due to internet booming-commerce and blogs where by on cessation of business, collected personal information of people is sold and used in unethical way. Security issue is another threat of data mining, whereby hackers access and steal both personal and financial information of employees and customers and use that information in fraudulent activities. In addition to that, misuse of information is another drawback of data mining whereby unethical businesses exploit information for personal benefits.

Data mining algorithm is a process of manipulating data from database using heuristics and calculations that create a model from data. Examples of data algorithm examples include C4.5 which is used to construct a decision tree when fed with classified data.C4.5 also works on both continuous and discrete data through specifying ranges for continuous data which converts continuous data into discrete data.C4.5 is mostly used in an open source data visualization and analysis tool.it has the advantage of producing output that is human readable in a decision tree because of ease of interpretation and explanation.

K-means is another example of data mining algorithm which is used to create groups that are more similar versus non similar groups. For example using a dataset of library books with distinguishing characteristics like biology books, magazines, business books and chemistry books, k-means can classify books according to their subjects for example science books which comprises both biology and chemistry books.it is designed to only work on continuous data it is mostly applied in MATLAB and k-means has the advantage of being faster and efficient than other types of algorithms Its sensitivity to outliers disadvantage it.

Support vector machine is also an example of data mining algorithm which learns mathematical function equations to divide data into two classes using a hyper plane that splits the two classes and data points. It is mostly used in MATLAB and scikit-learn. It has the advantage of creating easily explainable points. Its disadvantage is its strenuous interpretation of results.

In addition apriori is another example of data mining algorithms which explores correlation among variables in database. It works by joining, pruning and repeat of database. It is used in Orange and AR tool. It has the advantage of ease of implementation and understand ability. Its major weakness is that it requires a larger memory space and a lot of time.

In addition PageRank is an example of data mining algorithms which is used to determine the relative importance of some object linked within a network of objects. Google search engine is an example of page rank which explores association of items on a web page.

There data mining algorithms alternatives like harsh trees and tries data structure. Trie data structure uses a prefix as a representation during retrieval of data and consists of nodes and edges. It is used to solve a lot of problems especially for information stored in strings in multiple memory strings. It has the advantage of being simple scalable and faster. It only disadvantage is that it does not work on single string information.

Hash tree is another alternative of data mining algorithm, it is a data structure used mainly in harsh values whereby descendants of each node is interpreted in memory. Its major weakness is its slow speed.

R programming is also an example of data mining algorithm alternative, it is an effective statistic software which is used in data analysis where it provides statistical and graphical technical skills for example in linear and nonlinear modelling and time series .

REFERENCES

Fong, J., & Siu, B. (1997). Data mining, data warehousing & client/server datab...