Mixing Scores from Artificial Neural Network and Social Network Analysis to Improve the Customer Loyalty

摘要

Due to the increased competition in the telecommunications, customer relation and churn management is one of the most crucial aspects for companies in this sector. Over the last decades, researchers have proposed many approaches to detect and model historical events of churn. Traditional approaches, like neural networks, aim to identify behavioral pattern related to the customers. This kind of supervised learned model is suitable to establish likelihood assigned to churn. Although these models can be effective in terms of predictions, they just present the isolated likelihood about the event. However these models do not consider the influence among the customers. Based on the churn score, companies are able to perform an efficient process to retain different types of customer, according to their value in any corporate aspects. Social network analysis can be used to enhance the knowledge related to the customers' influence in an internal community. This new proposition to valuate the customers can arise distinguishes aspects about the virtual communities inside the telecom networks, allowing companies to establish more effective action plans to enhance the customer loyalty process. Combined scores from predictive modeling and social network analysis can create a new customer centric view, based on individual pattern recognition and community overview understanding. The combination of scores provided by the predictive model and the social network analysis can optimize the offerings to retain the customer, increasing the profit and decreasing the cost assigned to the marketing campaigns.