As customers use and benefit from multiple services, a large amount of customer data are accumulating daily. Connecting a customer's identity on a service with her identity on a different service, known as user identity linkage (UIL), enables a comprehensive understanding of users in a variety of real-world applications. The difficulties of UIL tasks in marketing applications are mainly the lack of user demographics and diverse user behavioral patterns, which differs from UIL tasks in social networking services that previous UIL methods have mainly been used to tackle. In this paper, we propose a novel method for UIL for different behavioral patterns to determine whether two given behavioral histories come from the same user without using any user demographics. Our proposed method links users by using natural language processing to efficiently characterize user intrinsic features and bridging the gap between two different behavioral patterns of the same user. We conducted experiments to evaluate our proposed method for three real-world open source datasets and observed that it successfully linked users compared to conventional UIL methods.