Tourism Activity Discovery And Tourist Behavior Mining System Based On Spatio-temporal Context

With the rapid development of mobile Internet and computing capabilities, it is more and more important to use artificial intelligence to discover the popular tourism activities and analyze the tourist behavior of tourists. In order to better identify and discover tourism activities and analyze tourist behavior patterns, this thesis proposes an algorithm for tourism activity identification and discovery based on large tourism data in the context of time and space. At the same time, the behavior of tourists in the context of time and space are deeply excavated. The main work of this thesis is as follows:(1) This thesis proposes an improved feature-based dimensionality reduction algorithm for tourism data collection and analysis based on time-space context. The traditional CHI value calculation is improved with weighting the positive and negative correlations of the category. The proposed algorithm extracts feature via pattern aggregation theory,considering the the influence of feature on the contribution degree of category. It also merges items that have high classification similarity, and re-select feature with the decisional table based on rough sets. Compared with the original travel algorithm, the improved ICAPC algorithm is 5.26%higher than the original algorithm in F1-score performance. It demonstrates that the improved feature dimensionality reduction algorithm proposed in this thesis classify the text more accurate.(2) This thesis proposes a tourism activity discovery method based on TATM model for tourism activities recognition and discovery based on space-time context. With the topic probalility of travel, the method classifies the travel information of tourism activities and mines the potential of tourism activities. With the tourism activity probability of the topic, it discovers the distribution of each tourist activity and intuitively find the corresponding theme of each tourism activity. In the F1-score of tourism activity classification, the proposed TATM is 11.29% higher than the tourism activity classification based on SVM, 21.32% higher than that of Bayes-based tourism. It proves that the TATM proposed in this thesis can effectively identify and discover tourism activities.(3) This thesis proposes an association rules mining algorithm for temporal behavior patterns (ARMTBP). This method clusters the time series behavior of visitors and obtains the sequence pattern of tourists'behavior. Then it extracts the association rules of the visitors' timing behavior patterns and extracts the tourist behavior patterns via mining the frequent itemsets. This thesis conducts experiments to measure the performance of the algorithm by comparing the JM value of the rule pattern set. Experimental results show that when the optimal cycle size is w = 6, the JM value is 6.92% higher than that of w = 4. Under the optimal JM value, the confidence level of the visitor timing behavior pattern rule also increases by 5%. They demonstrate that the visitor behavior pattern mining algorithm based on the spatio-temporal context can effectively obtains the visitor behavior pattern.(4) This thesis designs and implements a tourism activity discovery and behavior pattern mining system based on space-time context. The system realizes the functions of tourism data collection and analysis,tourism activity identification and discovery, and tourist behavior pattern analysis. The system also compares the benchmark dataset with other algorithms to prove the validity of proposed methods. Testing results prove that the system completed the corresponding function, and the running results are as expected.