Algorithm 1. The processing of generating behavioral semantic clustering.

Input: semantic features $\mathrm{F}=\left\{f_1,f_2,\dots ,f_m\right\}$

The number of category $k$

Output: $S=\left\{S_1,S_2,\dots ,S_k\right\}.$

Repeat

for $j=1,2,\dots ,m$ do {

According to, calculate the posterior probability of $f_j:$ ${\lambda }_{ji}=p_M\left(z_i=i|f_i\right)$}

for $i=1,2,\dots ,m$ do {

Calculate mean vector: ${\mathbf{\mu }}_{\mathit{i}}=\frac{{{\displaystyle \sum }}_{j=1}^m{\lambda }_{ji}{f}_j}{{{\displaystyle \sum }}_{j=1}^m{\lambda }_{ji}}$

Calculate covariance matrix: ${\Sigma }_{\mathit{i}}=\frac{{{\displaystyle \sum }}_{j=1}^m{\lambda }_{ji}{\left(f_j-{\mathbf{\mu }}_{\mathit{i}}\right)}^{\mathrm{T}}{\mathsf{\Sigma }}^{-1}\left(f_j-{\mathbf{\mu }}_{\mathit{i}}\right)}{{{\displaystyle \sum }}_{j=1}^m{\lambda }_{ji}}$

Calculate mixture coefficient: ${\alpha }_i=\frac{1}{m}{\displaystyle \sum }_{j=1}^m{\lambda }_{ji}$}

Update semantic parameters $\left\{\left({\alpha }_i,{\mathbf{\mu }}_{\mathit{i}},{\Sigma }_{\mathit{i}}\right)|1\le i\le k\right\}$

Until find out the optimization of ${\lambda }_j$

for $j=1,2,\dots ,m$ do {

According to ${\lambda }_j=\underset{iϵ\left\{1,2,\dots ,k\right\}}{\arg \max }{\lambda }_{ji}$, generate behavioral semantic clustering:

$\mathrm{S}={\mathrm{S}}_{{\lambda }_j}\cup \left\{f_j\right\}$

}