Study On Passenger Preferences Mode In Airline Passenger Personalized Service

This thesis, personalized service as the center, studies personalized recommendation algorithm to find passenger's travel preferences mode based on passengers booking records.First, a brief introduction of the three commonly used personalized recommendation algorithm is given, including content-based recommendation algorithms, collaborative filtering recommendation algorithm and hybrid recommendation algorithm.On this basis, two improved recommendation algorithm are proposed. One is based on resource allocation user similarity to improve the Slope One algorithms. Too much reliance on the number of evaluation projects in the original algorithm results in user weak pertinence and predicted deviation. A resource allocation user similarity measurement method is proposed. Increase the weight of the similar user to improve Slope One algorithm, and test on standard data sets Movie Lens. The test results show that the improved algorithm increases the prediction accuracy to some extent. Second,the improved bipartite resource allocation algorithm is proposed to improve the resource allocation matrix, and to coupled with the original matrix to improve the recommendation accuracy. The accuracy of improved method changes with the variable parameters. By testing on standard data sets Movie Lens we find that when the variable parameters have their optimal values, the accuracy of improved method is superior to the former unmodified.Finally, use the two improved algorithms to find passenger travel preferences mode. Comparing the results of application with the original data set, we indeed discovered the discipline of passenger travel. And then compare the two improved algorithms from different angles.