MSS - Report Assessment : Techniques and Strategic Application of Data Mining in Insurance Industry

MSS - Report Assessment

Case Scenario/ Task

Write a report to answer the following questions. See the marking rubric for the marking schedule.

Question 1

Analyse and evaluate (using ‘Asking Questions’ handout) how the Data Mining technique can or might be used in an industry of your choice: Focus on the technique and how it can be used by the industry in general rather than a particular organisation.

Question 2

For an organisation of your choice, identify an existing or potential business strategy for that organisation and analyse (using ‘Asking Questions’ handout) the role that ‘Big Data’ could play in meeting the strategy’s objectives.

Introduction

(a) Purpose of investigation and significance In the current times, the insurance industry is constantly expanding due to increase in population as well as risks associated with the daily life activities. The insurance companies

believe in gathering the maximum data that is further processed and help in increasing their business opportunities. The series wise process of extracting useful information from large data base is called data mining, which has become the most essential activity of insurance companies these days. The following research has been conducted in order to gain better understanding of the significance of data mining and necessary techniques to identify the patterns in huge data.

With the help of advance technology and knowledge, companies evaluate and analyse the patterns or data, and frame policies accordingly. On the basis of the analysed data, various plans are formulated and offered according to the basic requirements of the insured. Data mining has made it easier for insurance firms to manage huge data and spot the useful information.

(b) Research method

The research has been conducted on 5 employees of varied insurance companies. The information regarding the techniques and algorithms of data mining have been gathered through primary method of data collection. All the five employees have been thoroughly interviewed on various topics associated with data mining, its methods, significance and practical implications in the operations of insurance companies.

(c) Impact of any limitations imposed on the report The interview method for collecting primary data offers more flexibility to the information collected for the purpose of the research work. It facilitates clarification of topics and queries, which is vital for the explanation of the research topic. However, it is observed that interview method takes more time and cost as compared to other methods of data collection.

Most times, interview method takes relatively more time to extract data and useful details from respondents, which leads to low response rate (McNabb, 2008). Moreover, the efficiency of the interview depends upon the knowledge, skill and understanding of the interviewer about the concerned topic. It is worth emphasizing here that the interpretation of data and findings of the research work are highly influence by the personal understanding of the interviewer. Few researchers believe that interview method of data collection is highly subjective as the perspectives of interviewers can vary even on similar topics (McBurney and White, 2009). Every interview carries a distinct conversation which makes it challenging to compare the findings and data collected.

(c) Explanation of organisation of material Through the process of data mining, useful information is extracted from the pool of data. The insurance companies use this technique to identify the patterns in the market and needs of the insurance policies. This information proves to be the basis of further business development and formulation of strategies (Ramageri, 2012). The step wise implementation of data mining process helps in breaking down the complex information. In the process, firstly the data is explored and transformed in other forms so that thenature of information can be evaluated. After this, patterns are identified in order to make

(D)predictions and increase business prospects in the given scenario. At the last step, the identifiedpatterns are used for deploying various plans and policies of insurance with an intention toachieve desired output (Ramageri, 2012). There are various techniques of data mining asclassification, predication, association rule and neutral network.

(e) Definitions if needed

I. Data Mining : The logical process of searching through large amount of data to obtain useful and productive information.

II. Classification Trees: Classification tree mentions the choices available to the decision maker, where each possible decision and its estimated outcome are shown as a separate branch of the tree.

III. CART (Classification and Regression Trees): These are the methods used for predicting constant depending variables and determine the set of various logical conditions that can derive accurate predictions and classification of cases. the approach facilitates simple results, with breaking down the complex structure of the problem or the case

Findings

From interviewing five employees from insurance companies, it is learned that data mining is an integral part of the work process of insurance companies. Companies considerably rely on the information extracted through data mining for formulating insurance plans anddeveloping business activities. All five respondents, who are employed in various organisations, have considerably shed light on the methods used for data mining and their implications in the long run.

When questioned, two respondents mentioned that they use clustering technique for data mining as it helps them to identify the overall distribution pattern and correlations among data attributes. In order to be more cost efficient, clustering can be done as a prior activity for further classification of data. One of the employees also said that they form groups of customers on the basis of the purchasing patterns and likewise functionality. From further investigation, it islearned that the insurance companies employ partitioning method, hierarchical agglomerative method, density based method, grid based method and model based method for clustering of data.

Majority of respondents stated that classification is the most commonly used style for data mining that involves pre classified sets for developing models for classifying the population.

The technique is usually used for fraud detection and credit risk applications, and adopts decision tree algorithm. One of the respondents also indicated towards using various models of classification in order to achieve better results. In the insurance industry, it is very difficult to make predictions regarding stock prices and sales volumes as these elements are based on several complex variables. One of respondents stated that his company uses techniques like logistic regression and neutral nets to forecast the values for future. Most companies in the industry employ classification and regression model in order to predict constant response variable in the market. Linear regression, multivariate linear regression, non linear regression and multivariate non linear regression are basic types of regression analysis as explained by two of the five respondents.

From the interviews, it is also learned that while making decisions regarding the catalogue design, cross marketing and customer shopping behaviour analysis, companies use association rule, which can be multilevel, multidimensional or quantitative. Other than these methods, neutral network is also used as a data mining technique, which helps in deriving meaningful information from the complicated data, and extracts complex patterns and trends for forecasting requirements.

However, it is also found that the data mining technique still has a scope of improvement, but companies are using it on continuous basis. Most companies employ a mix of various mentioned techniques for data production in order to learn more about customer behaviours and smart marketing decisions. During the interviews, all five respondents provided significant

information in concern to data mining techniques applied in their respective organisations. The findings vitally suggest that data mining is a useful technology, which contributes to the effective decision making and operational activities of organisations in the insurance industry, where large

pool of data is involved and extraction of useful information is complex.

From the research and findings, it is learned that clustering is a common technique employed for data mining in the insurance industry. Clustering helps in summarizing the obtained data by recognizing the common features and characteristics. The technique is of great use in segmenting the targeted database and forming clusters for further evaluation. Most organisations in the insurance industry focus on implementing clustering data mining technique in order to recognize groups and analyzing the distribution and patterns in the data. The technique basically divides the data in similar groups, like segmenting the existing policyholders and associating a distinct feature to each group for future strategies. For data mining, Eucilidean distances are derived through quantitative variables and seeds are identified with the help of algorithms. These distances between the data observation and seeds are

measured through the cluster criterion, where all observations are grouped in clusters.

One of the basic features of clustering is that segmentation is done without any prior knowledge regarding the class of specific trait of the data. The homogenous classes are grouped from the huge database with an unidentified target. There are various techniques used by insurance companies in order to segment the data like pattern recognition, neutral nets,

COBWEB, Bayesian approach, etc. The conceptual clustering algorithms are used to evaluate the distinct features of all the records and classify each cluster in order to develop concept. These concepts connect the

attributes and their values in order to obtain the required base. By using the Bayesian approach, the most likely data are clustered through the algorithm. Gathering and evaluating the obtained data is a tedious job, which is simplified and can be performed with much ease with the help of data mining, especially in the insurance industry AAMI is one of the top insurance companies in Australia that provides personal insurance, business insurance and also covers personal and professional property. The company is a trusted insurer of more than 60,000 businesses and 2.8 million people in the country. Being a

vast organisation, it employs the most efficient and effective strategies and planning in order to operate at such vast level. It is a big challenge for insurance companies to meet the customer requirements and to be affordable at the same time.

It is observed in the recent times that people have a grown interest and necessity of insurance, which arise the need of innovate insurance deals, ideas and business policies for companies to sustain in competition as well as serve customers with better products. Data mining is an essential activity of insurance companies, which focuses on extracting the useful data out of

the large amount of information available. AAMI also uses this technique where with the help of knowledge discovery of data, it identifies the pattern of data that was unrecognised earlier. The company most often employs clustering technique of data mining where groups are formed on the basis of consumer behaviour. AAMI uses various clustering approaches like Partitioning method, hierarchical method, density based method, grid based method and model based method, depending upon the nature of the data and results required. Besides these methods, data mining is also done by regression, association, decision trees or classification approaches, which helps in extracting the most useful and optimum information from the pool of data.

Conclusion

From the above discussion, it is concluded that data mining plays vital role in developing business and forecast future values. In the insurance industry, companies employ various data mining techniques like classification, clustering, predication, association rule and neutral network in order to facilitate the search of patterns and take decisions for the future business

trends. Data mining essentially extracts necessary information and identifies patterns, forecast and discovery of knowledge in varied business scenarios. Various steps of data mining include exploration, pattern identification and deployment, which help in gathering imperative information for further use. The extracted information can be used for multiple purposes like marketing models, stock prices, offers and more. The research also indicates that most

organisations in the insurance industry employ clustering technique for data mining. The top insurance company of Australia, AAMI also performs the task of data mining through clustering strategy very often. However, various other approaches are also employed as per the need and nature of data.

It is worth emphasizing here that not only the insurance industry, but data mining is a popular technique in various other industries where large data is generated. With the help of advanced information technology, data mining has proved to be a useful application. Companies can enhance their productivity and grow business operations be applying the suitable method of

data mining. They can make suitable marketing, production and strategic decisions on basis of the derived information and predictions. In order to serve the fast changing consumer needs and achieve competitive advantage, companies need to adopt the most unique and productive approaches for their sustainability.

Recommendations

The research regarding the use of data mining techniques indicates that there are various techniques of data mining. All the techniques give both qualitative and quantitative results depending upon the nature of data and the business operations. However, it is recommended that companies in the insurance industry use the classification technique of data mining in order to achieve better outcome. The classification technique involves a set of models in order to develop a suitable framework for categorizing the population of records. With the help of classification data mining techniques, it is easier to detect fraud and credit risk applications (Plant and Murrell, 2007). As the technique generally employs decision tree or neutral based classification algorithm, it learning and classification are its integral activities.

In the data classification process, the training data is analysed by classification algorithm and the data vitally evaluates the accuracy of the classification norms in classification test. In case the accuracy is satisfactory, the classifications rules can be applied for fresh data. The technique is quite useful for companies as the fraud data application includes the analysis and

detailed evaluation of both fraudulent and valid activities identified on record. The classifier training algorithm of the classification technique considers the pre classified illustrations in order to evaluate the standards needed for effective discrimination. Further the standards are encoded into classifiers by the algorithm (Plant and Murrell, 2007). Moreover, there are various

classification models as classification by Decision Tree Induction, Bayesian Classification, Neutral Networks, Support Vector Machines and Classification Based on Association. The classification technique is more productive and alert in fraud detection, which helps in obtaining the genuine information and make decisions accordingly. From the entire research work, it is recommended that companies must employ classification technique of data mining in

order to garner the most productive information and data.