sitively correlated to satisfaction. They also took demographical factors (age, gender, education background and so on) as pa In our research we suppose that the satisfaction comes from the whole travel; and under agN Trudel of Huo and Miller budget and time the sum of each spots satisfaction. Here we referred to the satisfaction he expectation of travel can be described as Here "I"denotes a specific tourist spot, " f means the evaluation means the evaluation on services, and “e” means the evaluation on past expe n is the ncluding""representing the preference of visitors (effected by age, ger 品四= objective weights, and. This theory has already been validated and we will use it here as our standard to measure the performance of our advanced ACA [12] 4. System testing and analysis We implemented the prototype system according to the above designed functions, in order to see whether the ACA can indeed perform better in the aspect of user experience and preference when utilized in such context-a anning system. Functions for our prototype tour planning include user information input(age, number of visitor user setting key sites( that are sure to be visited), system recommending other sites according to users' preference, information abstracting, route optimizing, and multi-choice result presentation A. Testing: Some Computational Results In this section of our research, we used a real case of the Shanghai Zoo as a testing background; this would not only erve better in user satisfaction measurement, but also attach higher practical value to our advanced ACA and system unction design In order to verify the effici and effectiveness of the advanced ACA, the following two tests(Test 1, Test 2)are designed to gather information about its running condition under a specific scenario. They both focus directly on the rmance itself to get a reflection of the scientific rigorousness of our improvement. Test 1: Efficiency. We used a set of data to test whether the advanced al gorithm is able to get the optimized results in an acceptable time period, and whether it can calculate the best route in an efficient and convenient method Tab. I Testing Set and Results of Test l for efficiency Testing Set I of user L23457 2→>1→>4→>7→3->5→> 2→>1→>4→>7→3→>5→>8 nd Iteration I value or the route Testing Set 2 Analysis: In test 1, each test group was traced for 10 steps, and the program was all the time able to get the optimiz results under the constraints. The running times of all tests are around 100 ms quantitative level. At the same time, with mplexity of 15 spots involved, the program managed to provide a shortest tour 10 times of iteration In the 0. 182s solving time of test I, the program reached the final result on the third iteration at 0.037s. Test 2 took 0. 142s while calculating, and the optimized result appeared at 0.042s during the third time of iteration. Therefore, for tour planning matical scope, the advanced ACa system is effective and Test 2: Effectiver stem was tested under several specific scenarios to se er it can offer rational results lat satisfy the users have the ability to analysis and process the users' customized information in order to include proper sites in the re nded tour Tab 2 Testi Square, Pets World, Lion Mountain, MammalsTesting Set 1 User Input： Number of users 2 User ages 24,25 Key point sites 1,4,7 Total NO. of sites 7 System Output Recommend sites 2,3,5,8 All sites to visit 1,2,3,4,5,7,8 Testing Set 2 User Input： Number of users 4 User ages 13,15,37,39 Key point sites 3,6,9,12,13 Total NO. of sites 8 System Output Recommend sites 1,4,5 All sites to visit 1,3,4,5,6,9,12,13 positively correlated to satisfaction. They also took demographical factors (age, gender, education background and so on) as parameters to test and verify the correlation. In our research, we suppose that the satisfaction comes from the whole travel; and under a given budget and time constraint, the travelling process can be simplified as sequential visits to several separated spots. Therefore, the total satisfaction value equals to the sum of each spot’s satisfaction. Here we referred to the satisfaction model of Huo and Miller (2007), which has been verified and widely recognized. The expectation of travel can be described as: . (3) Here “i” denotes a specific tourist spot, “f” means the evaluation on facilities, “s” means the evaluation on services, and “e” means the evaluation on past experience. The expectation is the sum of these three values multiplied by weights, including “ ” representing the preference of visitors (effected by age, gender, profession and so on), subjective weight, and objective weights , and . This theory has already been validated and we will use it here as our standard to measure the performance of our advanced ACA [12]. 4. System testing and analysis We implemented the prototype system according to the above designed functions, in order to see whether the advanced ACA can indeed perform better in the aspect of user experience and preference when utilized in such context-aware tour planning system. Functions for our prototype tour planning include user information input (age, number of visitors, etc.), user setting key sites (that are sure to be visited), system recommending other sites according to users’ preference, context information abstracting, route optimizing, and multi-choice result presentation. A. Testing: Some Computational Results In this section of our research, we used a real case of the Shanghai Zoo as a testing background; this would not only serve better in user satisfaction measurement, but also attach higher practical value to our advanced ACA and system function design. In order to verify the efficiency and effectiveness of the advanced ACA, the following two tests (Test 1, Test 2) are designed to gather information about its running condition under a specific scenario. They both focus directly on the algorithm performance itself to get a reflection of the scientific rigorousness of our improvement. Test 1: Efficiency. We used a set of data to test whether the advanced algorithm is able to get the optimized results in an acceptable time period, and whether it can calculate the best route in an efficient and convenient method. Tab.1 Testing Set and Results of Test 1 for Efficiency Times of Iteration Sys Run Time (ms) Optimized route Total weighed value of the route 1 0 Start -> 1 -> 4 -> 7 -> 2 -> 3 -> 5 -> 8 - > End 95 2 20 Start -> 8 -> 1 -> 4 -> 7 -> 2 -> 3 -> 5 - > End 89 3 37 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 4 57 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 5 77 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 6 97 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 7 117 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 8 140 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 9 160 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 10 182 Start -> 2 -> 1 -> 4 -> 7 -> 3 -> 5 -> 8 - > End 88 Times of Iteration Sys Run Time (ms) Optimized route Total weighed value of the route 1 0 Start -> 4 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 5 -> End 124 2 22 Start -> 4 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 5 -> End 124 3 42 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 4 52 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 5 62 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 6 85 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 7 97 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 8 112 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 9 125 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 10 142 Start -> 5 -> 3 -> 6 -> 9 -> 12 -> 13 -> 1 -> 4 -> End 120 Analysis: In test 1, each test group was traced for 10 steps, and the program was all the time able to get the optimized results under the constraints. The running times of all tests are around 100 ms quantitative level. At the same time, with a complexity of 15 spots involved, the program managed to provide a shortest tour plan in 10 times of iteration. In the 0.182s solving time of test 1, the program reached the final result on the third iteration at 0.037s. Test 2 took 0.142s while calculating, and the optimized result appeared at 0.042s during the third time of iteration. Therefore, for tour planning problems with proper mathematical scope, the advanced ACA system is effective and efficient. Test 2: Effectiveness. The system was tested under several specific scenarios to see whether it can offer rational results that satisfy the users’ preferences in those cases. In such concrete situation, the algorithm must have the ability to analysis and process the users’ customized information in order to include proper sites in the recommended tour. Tab.2 Testing Scenario and Results of Test 2 for Effectiveness Scenario 2.1 User Input： System Output: Number of Visitors 2 Recommend Sites 9,5,15,12 Visitors’ Ages 10,11 All sites to be visited 9,5,15,12, 1,2,7,10,14 Key Point Sites 1,2,7,10,14 Sites Info：Panda, Kids’ Zoo, Glonde Park for kids, Marine Mammals’ Show, Science Edu Pavillion, Grand Show Square, Pets’ World, Lion Mountain, Mammals’ Povillion Number of Sites 9 Scenario 2.2 User Input： System Output: Number of Visitors 4 Recommend Sites 6,3,11,8,4