Mining the network value of customers
Loading...
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Одна из основных задач анализа данных - помочь компании определить потенциальных клиентов рынка. Если прибыль, которую может принести клиент, больше, чем стоимость маркетинговых действий, направленных на него, то маркетинг для этого клиент оправдан . До недавних пор работы в этой области учитывали только внутреннюю ценность клиента, то есть ожидаемую прибыль от клиента. Я предлагается рассматривать сетевое значение клиента, то есть ожидаемую прибыль от других клиентов, которым исходный клиент может предложить купить товар или повлиять на их выбор.
Решение основано на моделировании социальной сети в виде марковских случайных полей, где вероятность покупки каждого клиента зависит как от внутреннего желания приобрести продукт, так и от влияния других клиентов. В качестве практического приложения рассматривается общедоступная база данных MovieLens с рейтингами для 9 тысяч фильмов.
One of the main tasks of data analysis is to help the company identify potential market customers. If the profit that a client can bring is more than the cost of marketing activities aimed at him, then the client is justified in marketing. Until recently, research in this area took into account only the intrinsic value of the client, that is, the expected profit from the client. I propose to consider the network value of the client, that is, the expected profit from other customers to whom the original client can offer to buy a product or influence their choice. The solution is based on modeling a social network in the form of Markov random fields, where the probability of each customer’s purchase depends on both the internal desire to purchase the product and the influence of other customers. As a practical application, the publicly available MovieLens database with ratings for 9 thousand films is considered.
One of the main tasks of data analysis is to help the company identify potential market customers. If the profit that a client can bring is more than the cost of marketing activities aimed at him, then the client is justified in marketing. Until recently, research in this area took into account only the intrinsic value of the client, that is, the expected profit from the client. I propose to consider the network value of the client, that is, the expected profit from other customers to whom the original client can offer to buy a product or influence their choice. The solution is based on modeling a social network in the form of Markov random fields, where the probability of each customer’s purchase depends on both the internal desire to purchase the product and the influence of other customers. As a practical application, the publicly available MovieLens database with ratings for 9 thousand films is considered.