Based on Rough Set Theory, this research addresses the effect of attributes/features on the combination values of decisions that insurance companies make to satisfy customers’ needs. Attributes impact on combination values by yielding sets with fewer objects (such as one or two objects), which increases both the lower and upper approximations. It also increases the decision rules, and degrades the precision of decisions. Our approach redefines the value set of attributes through expert knowledge by reducing the independent data set and reclassifying it. This approach is based on an empirical study. The results demonstrate that the redefined combination values of attributes can contribute to the precision of decisions in insurance marketing. Following an empirical analysis, we use a hit test that incorporates 50 validated sample data into the decision rule so that the hit rate reaches 100%. The results of the empirical study indicate that the generated decision rules can cover all new data. Consequently, we believe that the effects of attributes on combination values can be fully applied in research into insurance marketing.
An understanding of customer consumption trends, and purchasing intentions and behavior is very important in sales and marketing. It is well known that customer satisfaction with products or services is an important key to achieving successful business operations and sustaining competition. Customer satisfaction is a critical issue in keeping customers, so measuring the impact of buying behavior on customer satisfaction is critical to understanding customer needs. Customers’ repurchase intentions and the retention of customers is driven by customer satisfaction (Kim, Ferrin, & Rao, 2003). The fulfillment of customer needs is related to satisfying customer expectations, which achieves customer satisfaction.
The insurance industry needs face-to-face contact with customers in order to provide services that satisfy customers’ needs so that they continue buying insurance, and/or to repurchase products. This is the main source of revenue for the industry. Due to its specialized nature, the insurance industry needs up-to-date information in order to modify products or services to attract potential customers. The best source of data is a market survey, the results of which may provide information about customers, such as their needs, purchase intentions, and service requirements.
We have designed a questionnaire about insurance products, purchase intentions, the budget for the premium, and participants’ basic data that may serve as the basis for understanding their needs. The consumers’ purchasing decisions and processes are analyzed, and proper marketing strategies and management operations are proposed. The results may be fully applied by managers to make decisions about strategies and processes related to consumer purchasing.
Most papers deal with insurance audits, purchase intentions, purchase channel studies, methodologies for investigating customer purchasing intentions, and customer satisfaction (Hennig-Thurau & Klee, 1997). Many research papers have quantified the problem in order to simplify the parameters, such as social parameters, and use statistical tools to analyze data. This approach, however, is only good for crisp types of data sets and certain data values. If the value of data is continuous or uncertain we must apply fuzzy theory (Zadeh, 1965).
Rough Set Theory is used in this study to analyze the content and features of data. The theory, which was developed by Pawlak (1982), is a rule-based decision-making technique that can handle crisp datasets and fuzzy datasets without the need for a pre-assumption membership function, which fuzzy theory requires. It can also deal with uncertain, vague, and imperceptible data. Until now, analysis of the attributes of combination values using Rough Set Theory has only been addressed by a few papers. In this study, a questionnaire with single-choice and multi-choice answers is used to apply Rough Set Theory to investigate the relationship between a single value and a combination of values of the attributes. Based on expert knowledge, the value class of the questions with multi-choice answers is reclassified in order to simplify the value complexity, which is useful in the decision-making procedure.
After the study, we applied a hit test to check the feasibility of the decision rules. As the hit rate reaches 100%, it is clear that new data fits the decision classes. The results of this research demonstrate customers’ insurance needs. The results are as follows: the purchase expectation is endowment; the age of target customers is 25–35 years; and the most purchased product is a mixture of products.
The remainder of this paper is organized as follows. Section 2 describes the methodology of Rough Set Theory. In Section 3 a real case of insurance marketing is presented to show the process of the effects of attributes on combination values. Finally, in Section 4, we present our conclusions.
This study demonstrates customers’ needs for the insurance market in Taiwan. From the results of survey, the following conclusions can be drawn: (1) the age group 25–35 is the highest insured target customer base; (2) the insurance purpose is endowment; (3) an acceptable annual premium is under NT$30,000 (US$938); and (4) the most purchased product is a mixture of products.
Multi-choice questions increase the number of elementary sets, so the expert’s contribution in the input data pre-processing task is important. We demonstrate that redefined attribute values can reduce the number of elementary sets and show that too many sets of one object will increase the approximation (classification) accuracy, but reduce the precision of decisions. In addition, every attribute value’s number can affect the class number, and multi-choice questions may have combination values. If the values of questions are unique, then reclassifying the results will not be beneficial. Here, we suggest that a hybrid system combining Rough Set Theory with Genetic Programming could be used to perform rule extraction, and may overcome the shortcomings of too many rules under the attributes of combination values. This will be addressed in our future work.
