Determining the target audience by data mining method in language educational institutions abroad

Today, with the development of technology and the innovations it brings, the rate of foreign language learning and its importance has increased significantly. Learning a foreign language in the global order has started to have a critical importance in professional business life, regardless of the sector. This importance has started to be seen more in private sector employees. Due to this importance arising from the private sector, knowing a language has been one of the most determining factors in recruitment. Although university students mostly prefer student exchange programs in terms of both cost and time, this situation has started to be insufficient. For this criterion of great importance, many overseas education consultancy companies have been established in Turkey, which can be used by individuals from all walks of life. These organizations implement programs that appeal to all ages, both educational and working, and traveling. The increase in the number of these organizations day by day has started to cause a competition among the organizations. In this study, it is aimed to learn which criteria affect the registration status by collecting the customer information of the consultancy company discussed in this study. According to this result, defining the customers, determining the target customer group, determining the customer profile that can be offered special offers, changing the customer portfolio when necessary, determining the times when they receive the most customers, it is aimed to present the necessary improvement suggestions. It is aimed to fill the gap in the literature by not finding a study in the sector-specific data mining approach discussed in the study. The data required for the study were brought together by bringing together the data that the education consultants kept when the customers came for information or when they called for information. The reason for considering the company, which is still new in the sector, aims to progress rapidly due to the fact that it is a new consultancy company in the sector compared to rival companies. Data of 727 customers were taken for the study, 8 independent variables and 1 dependent variable were included in the data set. While the dependent variable was about whether to register with the company or not, the independent variable was the information of those who registered or only received consultancy. Our independent variable factors are divided into: gender, age category, department, status, city sought, overseas experience, program applied for, and date sought. Gender variable was divided into two groups as male and female, age category has five groups within itself, the field variable is divided into eight groups, the status variable is divided into two groups, depending on whether the individual is a student or an employee. the city variable is divided into five groups within itself, into two groups according to whether the individual has experience abroad or not, the applied program is divided into five groups within itself and the last sought date is divided into 12 groups within itself. SPSS Modeler program was used in the study and C5.0 and C&R algorithms from decision tree methods were used. The C5.0 algorithm determines entropy values while distinguishing trees and distinguishes according to entropy values. The first method was made according to the variable area of the first distinction from the C5.0 algorithm model, followed by the searched date, gender and foreign experience criteria. In this case, according to the model output results, the most important factors affecting enrollment are; field, date sought and experience abroad. In particular, the common point of the cluster, which left 70% and 64% of the classification groups as registered, was the fact that the field of the individuals was outside of engineering, being a male individual, and not having experience abroad. When the results of the two algorithms are compared, it is noteworthy that the determining factors in both algorithms are the field, the city and the searched date. According to the analysis result of the C5.0 algorithm, when the performance value of the model was interpreted for the test data, it correctly interpreted 55 unregistered persons as unenrolled, and misinterpreted 31 unregistered persons as enrolled. It correctly classified 29 unregistered persons as not registered as not registered, while it correctly classified 65 unregistered persons as registered. In this case, while the model achieved 69.74% success for the training part, it achieved a 46.67% success rate for the test data. According to the C&R algorithm analysis result, when the performance value of the model is interpreted for the test data, it correctly interprets 51 people who did not register as not registered, misinterpreted 57 people who did not register as registered. It correctly classified 56 unregistered people as not registered as not registered, and correctly classified 52 unregistered people as registered. In this case, 57.22% success was achieved for the model training part, while a 50.23 percent success was achieved for the test data.