Makine Öğrenmesi Tekniklerinin Bütçe Verimliliğine Uygulanması Üzerine Bir Çalışma
A Study on the Application of Machine Learning Techniques to Budget Efficiency

Summary/Abstract: Purpose - This study aims to select and score the customer base that can be beneficial to the business for marketing purposes by using the large amount of sales and promotion data. Design/methodology / approach - For this purpose, the sales information collected under the heading "Breakfast at the FRAT", which was provided by Dunnhumby Company, which provides services to brands and retailers through customer data, for scientific purposes and obtained from weekly breakfast products, constituted the experimental data set of the study. An original model has been proposed to identify customers for whom marketing can be more effective and efficient by using XGBoost algorithm, which allows the marketing budget to be spent only on the potential customer rather than all its customers. Findings - The analyzed data cover a period of 156 weeks between 2011 and 2019. Within the scope of the study in which the number of features and complexity conditions are minimized, the criterion parameters of the model performance have high success rates. These rates reveal that the algorithm used for the model created to spend the budget to be used in marketing to the appropriate customer group is appropriate. Discussion - It is thought that the findings resulting from the analysis of big data with machine learning techniques will contribute to data science and the method followed in the study will form the infrastructure of a budget support system where businesses can make financial estimates and predictions. It is possible to contribute and guide data science by determining the classification algorithm with the best estimate in this study, aimed at increasing the efficiency of the budget to be allocated for marketing by using the effect of a few feature groups over the sales obtained and made from real world data. There is also the possibility of contributing to the business world by using a model in which the infrastructure of the study is further developed by businesses.

• Details
• Contents