Research On Fine-grained Tourists' Behavior Preference Based On Inverse Reinforcement Learning

Because of the development of China's economy,people have more ways of leisure and entertainment.Traveling has become an important way for people to pursue the quality of life.People usually look up information about attractions on the Internet and make detailed travel plans.With the rapid development of the tourism industry and the continuous development of attractions in various places,there are more large attractions.However,the relevant attractions information on the Internet can only be accurate to the level of scenic spot.So through browsing historical records of scenic spots information or rating data of scenic spots on the Internet can only obtain coarse-grained tourists' preferences,but not obtain fine-grained tourists' preferences inside attractions.Therefore,how to obtain the tourist's data inside attractions to learn the fine-grained preferences of tourists has become a current research hotspot.The popularity of smart phones has greatly facilitated people's daily lives,and with the rapid development of microelectronic technology,the sensors embedded in smart phones have become more diverse,and these sensors always record people's behavior data.Sensors such as cameras and accelerations embedded in smartphones can also reflect tourists' behaviors,and these behaviors often imply tourist preferences.Therefore,through the combination of mobile sensor technology and location awareness technology in smart phones,fine-grained tourists' behavior data of tourists can be collected and obtaining finegrained tourists' preferences of attractions.Aiming at the above analysis of current tourism research hotspots,the popularity of smart phones,and the diversity of embedded sensors,the main research contents of this article are as follows:(1)To address the problem that fine-grained data cannot be obtained for tourists inside the attractions,we have designed a tourist's behavior data collection application that based on Android system to collect fine-grained data for tourists' behavior.We arrange the scene of experimental data collection and obtain the location information from the Bluetooth low energy device i Beacon.Through acceleration sensor on smart phone to collect acceleration data of tourist behavior and then preprocess the acceleration data,using acceleration data to calculate the number of tourist's stays.Finally,we count the number of taking photos,stay time and other tourist behaviors of tourists.(2)In order to learn tourists' fine-grained preferences,we construct a Markov decision process based on the tourist's behavior that use the Markov decision process to simulate the tourist tour.Dynamic simulation of agent and environment,action,state,immediate reward,cumulative reward and other elements in Markov decision-making process is carried out to form the framework of tourist behavior preference learning.We construct a reward function and add tourist's behavior characteristics to the reward function from collected tourists' behavior data.Finally,we design Inverse Reinforcement Learning algorithm to learn finegrained preference based on the real data.