Secure Personalized Recommendation System for Mobile User - Information Security and Cryptology-ICISC 2010 - page 271

2 get user location
//retrieve user location from GPS satellite
3 get user profile
//retrieve user information from the user profile or create user profile for new
user
4 if (condition) then (action)
//take the action by the condition of Rule-Based Personalization
5 search query information from the database
//search the corresponding information from the database of knowledge
repository
6 compute similarity (identify the two items that are most similar)
//compute similarity using Pearson Correlation Coefficient
7 compute weight
//compute weight using Significant Weighting (eq.(1))
8 compute trust-weight
//compute weight using Trust Weighting (eq.(3))
9 compute prediction
//compute prediction using (eq.(2))
10compute MAE
//compute Mean Absolute Error (MAE)
11display result (recommendation)
//display the result to the user which is closet to their interest
End
Fig. 2. Modified RPCF Algorithm
3.2 Process of Modified RPCF Algorithm
Fig. 3 shows the recommendation process of modified RPCF algorithm. The
modification process made to RPCF algorithm are computing significant weighting
after computing similarity and trust weighting before making prediction.
As an example, rule-based filtering process takes the user's request and used the
hotel rating to make recommendation to the target user. In this example, hotel rating
profile contains the attacks inserted by the attacker and the ratings of the
neighborhood users. The collaborative filtering method firstly computes the similarity
values of the hotel rating for the targeted user among neighborhood users.
The item-based filtering method is used to filter the item according to the user's
request and user-based filtering is used to filter the most appropriate item among the
filtered items of the rule-based process based on the similar neighborhood users.
For instance, the user's request is “Hotels in Yangon”, the system accept the user's
request and retrieve the user's current location from the GPS and search the hotels
according to the user's current location by using rule-based filtering. After searching
the hotels, the system will give the most appropriate recommendation to the user
according to the user's interest by applying collaborative filtering which is computed
from neighborhood users.
During this computation process, the significant weighting is computed by using
the similarity values of the neighborhood users. According to the significant
weighting results the fewer commonly rated items are pushed out the neighborhood