Formulating Business Understanding of Data Mining Projects in Customer Support Domain

 Despite advances in the capabilities of data mining algorithms, the knowledge extracted by these algorithms require a great deal of refinement to become actionable for business executives. The main objective of this research is to identify common business problems of customer support domain that can be solved with the help of data mining techniques and to formulate an ontology of business understanding. Hence, customer support business problems are first identified through interviews with the domain experts, and then, with the review of related literature, the ontology of customer support problems is developed. As results of the research, the business objectives of the customer support domain that lead to value creation and profitability are identified along with key activities and outputs of each activity and the analytical steps required based on data mining techniques to achieve each business objective. Finally, based on CRISP-DM data mining model, the ontology of business understanding is introduced.

Customer support is the establishment of a set of processes and systems aimed at achieving customer satisfaction and loyalty and ultimately creating a profitable and lasting business relationship. Customer-related data and information technology tools are the foundation of a successful customer support strategy. Today, the rapid growth of information technologies and the Internet have opened up new opportunities to create profitable and successful customer relationship through data mining. Business decision makers expect the output of the data mining process to be easily interpreted and applied. Therefore, data mining techniques should provide actionable knowledge to these people. The lack of focus on business understanding in formulating the data mining problem and the lack of knowledge of the business domain is the root of this problem (Cao, 2017. Li, 2014). So far, data mining has been seen as a data-driven process that has little focus on the context and knowledge of the business domain. The main purpose of this article is to model key customer support business problems, goals, and activities that can be solved by data mining techniques. In this regard, ontology is used as a tool for structuring and displaying knowledge of customer support. The ontology is a detailed and transparent explanation of a common conceptualization (Gruber, 1993). Cao argues that if the ontology integration between the framework and the methodology of data mining and organization/business can be established, the actionability of data mining output can be assured.

This study presents a model and method that facilitates business understanding of customer support for data mining problem and actioable knowledge discovery. To this end, first by interviewing software support experts in the software industry, the key concepts that are essentially business problems addressed by data mining algorithms are identified and presented in the form of basic artifacts. In the next step, related literature has been reviewed to generalize the basic artifact to the field of consumer support domain. The final artifact has been reconciled with the help of experts. Accordingly, the interview questions were formulated by experts. 1-What are the key activities in the area of ​​customer support? 2- What are the main goals in customer support? 3-What kind of data is stored in customer support? The purpose of the above questions is to identify the leading business goals in the area of ​​customer support first. Then, to reach each goal, data analysis steps must be made on specific data, and finally the outputs of the steps identified. The following five queries were searched through Google Scholar for literature review. The reason for choosing Google Scholar is to index all the databases in the world by the search engine's search algorithm. The last ten years date range, between 2009 and 2019, was considered as the search time limit. - Customer relationship + Literature review - Customer support + Literature review - Customer service + Literature review - Customer experience + Literature review - Customer journey + Literature review The reason for the literature search query in the search query is that identifying relevant concepts in the domain of customer support and building the ontology of understanding customer support business as the research goal is pursued.

Customer support business objectives were identified. Each goal has analytical steps that must be applied to specific customer support data. For example, two analytical steps have been identified for the purpose of optimizing revenue from existing customers, including finding the relationship between features and needs and matching the service portfolio to customer needs. These analytical steps are meant to classify customers based on similar features and assign customers to the product based on similar features. For this purpose, background data and client transaction databases should be used. The formulated artifact of the research acts like this.

Most efforts to improve the data mining process have so far focused on improving data mining modeling, processes, and algorithms. There is a profound semantic gap between what a knowledge engineer deduces from data mining algorithms and what a business user pursues for profitability and business metrics. There are two major limitations to an approach that overcomes knowledge engineering: first, business requirements are not considered correctly, and second, the business user does not have the technical and statistical knowledge needed to perform the data mining process himself. Current ontologies of data mining are generally focused on the modeling and technical evaluation of data mining outputs. There is no evidence of a business-centric system modeling and architecture review among data mining ontologies. The artifact of this research has not only provided understanding of customer support business problems to begin the data mining process but also provided the basis for producing architectures for recording and sharing the extracted knowledge and production of a learning system. Future research should focus on the prototype production of a learning system that also characterizes the data mining process modeling step associated with any problem identified in the research artifact. In other words, data mining ontologies first need to be scrutinized, and then the mapping or integration between data mining ontologies and ontology business support ontologies defined in which the user identifies the problem and other stages of the data mining process can be proposed.