Abstract:Measuring the closeness of friendships is an important problem that finds numerous applications in practice. For example, online gaming platforms often host friendship-enhancing events in which a user (called the source) only invites his/her friend (called the target) to play together. In this scenario, the measure of friendship closeness is the backbone for understanding source invitation and target adoption behaviors, and underpins the recommendation of promising targets for the sources. However, most existing measures for friendship closeness only consider the information between the source and target but ignore the information of groups where they are located, which renders inferior results. To address this issue, we present new measures for friendship closeness based on the social identity theory (SIT), which describes the inclination that a target endorses behaviors of users inside the same group. The core of SIT is the process that a target assesses groups of users as them or us. Unfortunately, this process is difficult to be captured due to perceptual factors. To this end, we seamlessly reify the factors of SIT into quantitative measures, which consider local and global information of a target's group. We conduct extensive experiments to evaluate the effectiveness of our proposal against 8 state-of-the-art methods on 3 online gaming datasets. In particular, we demonstrate that our solution can outperform the best competitor on the behavior prediction (resp. online target recommendation) by up to 23.2% (resp. 34.2%) in the corresponding evaluation metric.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper attempts to solve the problem of measuring the closeness of friendship. Specifically, the paper focuses on how to quantify the closeness of friendship between two users in a social network. This problem has a wide range of application scenarios in practical applications. For example, online game platforms often hold activities to enhance friendship, in which users (referred to as sources) will only invite their friends (referred to as targets) to play games together. In this case, the measurement of the closeness of friendship is crucial for understanding the invitation behavior of source users and the behavior of target users accepting invitations, and can be used as a basis for recommending appropriate target users.
However, most of the existing methods for measuring the closeness of friendship only consider the information between the source user and the target user, while ignoring the information of the group they are in, which leads to unsatisfactory results. To improve this, the paper proposes a new measurement method based on the Social Identity Theory (SIT), which describes the behavior of target users tending to identify with other users in the same group. The core of the paper is to transform the factors in SIT into quantitative measurement indicators, which consider both the local information and the global information of the group where the target user is located.
### Specific contributions
1. **Propose a new TFC measurement method**: Based on the Social Identity Theory, a new Topological Friendship Closeness (TFC) measurement method is proposed, which can retain both the local and global information of the group.
2. **Extensive experimental verification**: Through comparative experiments with 8 state - of - the - art methods on three real - world online game datasets, the superiority of the proposed SIT - based TFC measurement method in predicting user behavior is proved.
3. **Online target recommendation application**: The proposed solution is applied to the online target recommendation task, achieving a significant performance improvement, with an improvement of up to 34.2%.
### Main steps
1. **Classification stage**: First, divide the candidate groups for each target user, and use Weakly Connected Components (WCC) to represent the local community structure of the user.
2. **SIT - based measurement definition**: According to the factors in SIT, measurement indicators such as User - Group Tightness (UGT) and Intra - Group Tightness (IGT) are defined.
3. **Tendency inference**: Use the XGBoost model to infer the identification tendency of the target user to each candidate group based on the SIT - based measurement indicators, and apply it to downstream tasks, such as behavior prediction and target recommendation.
### Key formulas
- **PageRank and personalized PageRank**:
\[
\mathbf{\Pi} = \sum_{t = 0}^{\infty}\alpha(1 - \alpha)^t\mathbf{s}\cdot\mathbf{P}^t
\]
where \(\mathbf{s}\) is the initial vector and \(\mathbf{P}\) is the transition probability matrix.
- **Group PageRank (GPR)**:
\[
\rho_C=\frac{1}{|C|-1}\sum_{v_j\in C\setminus\{v_t\}}\rho_j
\]
- **Group personalized PageRank (GPPR)**:
\[
\pi_{t,C}=\frac{1}{|C|-1}\sum_{v_j\in C\setminus\{v_t\}}\pi_{t,j}
\]
- **User - Group Tightness (UGT)**:
\[
\phi_{t,C}=\frac{\sum_{v_j\in C\setminus\{v_t\}}w_{t,j}\cdot\delta_{t,j}}{\sum_{v_j\in