Algorithm 1. Rating prediction based on sentiment for user $u_a$ and item $i_j$.
1.	Function sentiment_ratingPred (user $u_a$, item $i_j$) {
2.	//This function is used to obtain the$pr\_sen{t}_{aj}$term of Equation (1)
3.	//Step 1:
4.	FOR each item $i_k$ in the training set:
5.	IF user $u_a$ already rated item $i_k$ AND review score matches rating THEN
6.	Add $i_k$ to list of items $I$;
7.	//The result of this step is a set of m items  $I=\left\{i_1,i_2,\dots i_m\right\}$
8.	//Step 2:
9.	FOR each user $u_b$ in the training set:
10.	FOR each item $i_k$ in the set of items $I$:
11.	IF user $u_b$ already rated item $i_j$ AND user $u_b$ already rated item $i_k$       AND their review scores match ratings
12.	Add user $u_b$ to list of users $U$;
13.	//The results of this step is a set of n users  $U=\left\{u_1,u_2,\dots u_n\right\}$
14.	//Step 3:
15.	IF length(U)>0 THEN
16.	FOR each user $u_i$ in the set of user $U$:
17.	Compute $s_{a,i}$= sim (user $u_a$, user $u_i$) by applying cosine metric;
18.	Add $s_{a,i}$ to $S$;
19.	//The result is a set of n similarity values  $S=\left\{s_{a,i}\right\}$
20.	Set the K value to select the K nearest neighbors using S;
21.	Compute the predicted rating $p{r}_{aj}$ by applying the Equation (2);
22.	$pr\_sen{t}_{aj}=p{r}_{aj}$
23.	Return $pr\_sen{t}_{aj}$;
24.	ELSE
25.	Return 0;
26.	}