Journal of Convergence Information Technology Volume 5. Number 8 October 2010 where Wa=(wa, 14, wa, w)represents the best fit chromosome generated from a GA algorithm for user u, Finally, all the unrated items are sorted in non-increasing order with respect to the overall ratings and the first N items are selected as the Top-N recommended set 5. Experimental evaluation The proposed system is evaluated data collected from yahoo! movies web (http://movies.yahoo.com).InadditiontotheoverallratingeachYahoomovieprovidesanotherfour criteria:story, acting, direction, and visuals. All ratings have 13 possible letter grades ranging from F to A+. For our analysis, we change the letter scale to a numerical scale of 1 to 13, with 1 denoting the worst evaluation, grade F and 13 denoting the best evaluation, grade A+. We collect a total of 11109 rating data from 297 different users and 368 different movies. Each user rates at least 20 movies and each movie is rated by at least 20 users We employ the 5-fold cross-validation approach. First, we randomly divide the dataset into five groups. Then we run five rounds of tests, each time choosing one group of data as the test set and the ther four groups as the training set. The data in the training set is used as the cf database. For each data in the test set, we randomly select 10 non-zero entries and delete the rest. Our recommender system is then evaluated by comparing the Top-N recommendations it makes, given the test data, with the set of deleted items 5. 1. Evaluation Metrics We use precision and recall, two commonly used performance measures in the information retrieval community, to evaluate the quality of a recommendation. Precision is the fraction of recommended movies that the user really likes. Recall is the fraction of interesting movies that are recommended More precisely correctly recommended_movi recall= total movies_ by_users correctly recommended movie precision ded movies (10) However, there is al ways a trade-off between precision and recall. Increasing the number of recommended movies will reduce the precision and increase the recall. To balance both measures, we use another measure FI metric that gives equal weight to precision and recall and is given precision We compute each metric for each test user and the overall average values for the test data are taken measures of the quality of the recommendation 5. 2. Experimental Parameters In this section we define the parameters that will be used during the experiments as follows Recommendation number(RN): 3, 5, 10, 15, 20, 25, 30Journal of Convergence Information Technology Volume 5, Number 8, October 2010 ∑ ∑ = = × = 4 1 4 0 1 k k a k k ja k a ja w pw p , , (8) where ),,,( 4321 = wwwwW aaaaa represents the best fit chromosome generated from a GA algorithm for user ua . Finally, all the unrated items are sorted in non-increasing order with respect to the overall ratings and the first N items are selected as the Top-N recommended set. 5. Experimental evaluation The proposed system is evaluated using data collected from Yahoo! Movies Web site (http://movies.yahoo.com). In addition to the overall rating, each Yahoo movie provides another four criteria: story, acting, direction, and visuals. All ratings have 13 possible letter grades ranging from F to A+. For our analysis, we change the letter scale to a numerical scale of 1 to 13, with 1 denoting the worst evaluation, grade F and 13 denoting the best evaluation, grade A+. We collect a total of 11109 rating data from 297 different users and 368 different movies. Each user rates at least 20 movies and each movie is rated by at least 20 users. We employ the 5-fold cross-validation approach. First, we randomly divide the dataset into five groups. Then we run five rounds of tests, each time choosing one group of data as the test set and the other four groups as the training set. The data in the training set is used as the CF database. For each data in the test set, we randomly select 10 non-zero entries and delete the rest. Our recommender system is then evaluated by comparing the Top-N recommendations it makes, given the test data, with the set of deleted items. 5. 1. Evaluation Metrics We use precision and recall, two commonly used performance measures in the information retrieval community, to evaluate the quality of a recommendation. Precision is the fraction of recommended movies that the user really likes. Recall is the fraction of interesting movies that are recommended. More precisely, total movies usersbyliked correctly recommende moviesd recall ____ _ _ = (9) total recommende moviesd correctly recommende moviesd precision _ _ _ _ = (10) However, there is always a trade-off between precision and recall. Increasing the number of recommended movies will reduce the precision and increase the recall. To balance both measures, we use another measure F1 metric that gives equal weight to precision and recall and is given as recall precision recall precision F + ×× = 2 1 (11) We compute each metric for each test user and the overall average values for the test data are taken as measures of the quality of the recommendation. 5. 2. Experimental Parameters In this section we define the parameters that will be used during the experiments as follows: - Recommendation number (RN): 3, 5, 10, 15, 20, 25, 30 131