Nowadays, massive data has been brought by the rapid development of technology. When finding whether trajectory to be detected is abnormal under the premise of given normal trajectories, we innovatively propose 1) Seq2Seq model based on LSTM prediction network for trajectory modelling (SL-Modelling), and 2) abnormal trajectory detection method with spatio-temporal and semantic information. Firstly, SL-Modelling is used to obtain sequence-type trajectory models of normal trajectory groups directly for subsequent detection with no need to extract a large number of features manually and adapting to different sequence length. Then we introduce the concept of distance and semantic interest sequence that makes full use of spatio-temporal and semantic information of trajectories. Finally, the similarity between models and trajectory to be detected is calculated to detect abnormal trajectory. The experimental results of publicly available flight data set show that trajectory models obtained are descriptive enough to represent normal trajectory groups well, and the accuracy of modelling is higher than the existing advanced methods. Besides, the detection with spatio-temporal and semantic information has been verified that it has stronger detection ability with higher accuracy and takes less time.

In Node2Vec, the global structure of the network is neglected and the stochastic gradient descent (SGD) method is easy to fall into local optimum. Based on this algorithm, an improved link prediction algorithm combining machine learning and hierarchical representation learning for network (HARP) is proposed. This method first uses adaptive learning optimizer Adam instead of SGD to improve Node2Vec, then divides the nodes and edges of the original network graph into a series of smaller layered graphs by merging them according to HARP, and then uses the improved Node2Vec algorithm to extract features continuously, so as to realize network embedding. Finally, a social network link prediction model based on machine learning and HARP is established. A series of social network link prediction experiments are carried out. The results show that the new method has excellent performance and stability.

Considering the successful application of deep reinforcement learning (DRL) on tasks of moving objects, this paper innovatively applies deep deterministic policy gradient algorithm (DDPG) to complete the cognition task on multi-dimension and continuous communication emitter motion behavior. First, we propose a DDPG-based behavior cognition algorithm (DDPG-BC). It chooses direction, velocity, and communication frequency as state space, gains experience from interaction between network and environment and outputs deterministic cognition results. Second, under the condition of sufficient prior information such as geographic information, we further propose a novel algorithm named DDPG-based behavior cognition with Attention algorithm (DDPG+A-BC). It introduces attention mechanism into DDPG-BC which limits exploration scope and the randomness of initial state and improves the exploration efficiency and accuracy. The simulation experiments verify that DDPG-BC and DDPG+A-BC show good cognition ability on two different data set. And the algorithms are all superior to other DRL algorithm and existing cognition method with higher cognition accuracy and less time. In addition, we also discuss the influence of episode, reward function, and added attention mechanism on algorithm performance.