Discerning individual interests and shared interests for social user profiling

Abstract

Traditionally, research about social user profiling assumes that users share some similar interests with their followees. However, it lacks the studies on what topic and to what extent their interests are similar. Our study in online sharing sites reveals that besides shared interests between followers and followees, users do maintain some individual interests which differ from their followees. Thus, for better social user profiling we need to discern individual interests (capturing the uniqueness of users) and shared interests (capturing the commonality of neighboring users) of the users in the connected world. To achieve this, we extend the matrix factorization model by incorporating both individual and shared interests, and also learn the multi-faceted similarities unsupervisedly. The proposed method can be applied to many applications, such as rating prediction, item level social influence maximization and so on. Experimental results on real-world datasets show that our work can be applied to improve the performance of social rating. Also, it can reveal some interesting findings, such as who likes the “controversial” items most, and who is the most influential in attracting their followers to rate an item.

Appendix

Before conducting the experiment methods, we pre-processed the ratings as follows: let L R a t e and H R a t e denote the Lowest and the Highest rating in the training set, then

• $r^{1}_{u, v} = \frac {r_{u, v} - LRate}{HRate - LRate}$;

• $r^{2}_{u, v} = r^{1}_{u, v} - r^{1}$, where $r^{1} = \frac {1}{|{\Omega }|} {\sum }_{\Omega } r^{1}_{u, v}$;

• let $\{r^{2}_{u, v}\}$ be the input of the methods.

When we apply the trained models to the applications, we recover the predicted values as follows:

$$ \tilde{r}^{1}_{u, v} = \left\{\begin{array} {rl} 0, & \tilde{r}^{2}_{u, v} < 0, \\ 1, & \tilde{r}^{2}_{u, v} > 1, \\ \tilde{r}^{2}_{u, v}, & else. \end{array} \right\} \tilde{r}^{2}_{u, v} = r^{1}_{u, v} + \mathbf{u} \cdot \mathbf{v}. $$

(26)

$$ \tilde{r}_{u, v} = LRate + \tilde{r}^{1}_{u, v}(HRate - LRate) $$

(27)

$$ \varphi_{u, v} = \left\{\begin{array} {rl} 0, & \varphi^{1}_{u, v} < 0, \\ L_{u}, & \varphi^{1}_{u, v} > L_{u}, \\ \varphi^{1}_{u, v}, & else. \end{array} \right. $$

(28)

where $\varphi ^{1}_{u, v} = \mathbf {\rho }_{u} \cdot \mathbf {v} + L_{u} \cdot r^{1}$, $L_{u} = {\sum }_{u^{\prime } \in \mathcal {D}(u)} \frac {||\mathbf {w}_{u^{\prime }, u}||_{1}}{K}$ (||⋅||₁ means the l ₁ norm of vectors).

$$ \eta_{u, v} = \left\{\begin{array} {rl} 0, & \eta^{1}_{u, v} < 0, \\ T_{u}, & \eta^{1}_{u, v} > T_{u}, \\ \eta^{1}_{u, v}, & else. \end{array} \right. $$

(29)

and $\eta ^{1}_{u, v} = \mathbf {t}_{u} \cdot \mathbf {v} + T_{u} \cdot r^{1}$, $T_{u} = {\sum }_{u^{\prime } \in \mathcal {D}(u)} T_{u^{\prime }, u}$.

$$ \theta_{u, v} = \left\{\begin{array} {rl} 0, & \theta_{u, v}^{1} < 0, \\ P_{u}, & \theta_{u, v}^{1} > P_{u}, \\ \theta_{u, v}^{1}, & else. \end{array} \right. $$

(30)

where $\theta _{u, v}^{1} = \mathbf {p}_{u} \cdot \mathbf {v} + P_{u} \cdot r^{1}$, $P_{u} = {\sum }_{u^{\prime } \in \mathcal {F}(u)} \frac {\mathbf {h}_{u, u^{\prime }} \cdot \mathbf {w}_{u, u^{\prime }}}{K}$.