نويد عباسي

Customers are the most important asset of an organization. Without customers who remain loyal and develop relationships with the organization, there is no hope for an organization to survive and grow. That is why an organization must adopt a specific strategy to retain customers and strengthen relationships with them. One of the main activities to achieve this goal is to understand the unique behavior of customers because providing personalized services is much more effective and efficient than mass marketing. In order to achieve this goal, organizations use customer segmentation, which is one of the main branches of customer relationship management. Due to the increasing growth of information, without the use of modern data mining and machine learning methods, there will be no way to segment customers. In other words, traditional methods in this field are no longer effective and are obsolete. In this study, a hybrid two-stage model is presented to predict the customer cluster. It should be noted that the model of this study is implemented based on the Crisp method. At first, a model based on the developed RFM with 6 variables is built from the primary data. In the first stage, customers are divided into different clusters by several different algorithms. Then, in order to predict the cluster, the output of the first model is entered into the cumulative learning algorithm. The main goal of this study is to find the best combination of these two models based on accuracy eva‎luation indices. In order to eva‎luate more accurately, the innovative model based on Euclidean distance is used. Finally, the combination of K-Means and XGBoost algorithm is selected as the best combination with 78% accuracy. 6 categories of this clustering are also fully analyzed. Using a cumulative learning model instead of multiple clustering has higher speed and accuracy. Also, the use of this model prevents multiple analysis of clusters. This allows the customer to be placed in his category as soon as he buys from the organization, and marketing operations, especially personalized services, are provided to him from the very beginning. Also, these activities increase customer satisfaction and loyalty, and in today's competitive market, it can help to make that customer permanent for the organization. Naturally, the use of new models such as the one presented in this project will make the organization more dynamic, and as a result, it will lead the organization towards its main goal, which is to generate more income.

داده كاوي , يادگيري ماشين , خوشه بندي , يادگيري تجمعي , تقسيم بندي مشتريان

data mining , Machine Learning , clustring , Ensemble learning , Customer Segmentation

Navid Abbasi

Dr. Mahdi Ghazanfari