<p>This paper proposes a semantic analysis to pursue indoor trajectory, and include context-aware text mining to recognize indoor behavior. Based on RFID and POS data of supermarket, we consider customer visits of supermarket as indoor case study, and investigates impact of customer indoor trajectory on shopping behavior. Firstly, in acquisition of trajectory data, we carry out indoor experiment in a middle size of supermarket, and collect trajectory data via RFID tags which are attached to shopping carts. Secondly, in semantic analysis of trajectory, we transform customer trajectory into visiting sequences according to supermarket layout, where the sequences can be considered as contextual texts, and transformed into featured vectors by using Word2Vec approach. Finally, in pattern recognition of trajectory, we employ text-based convolutional neural network (TextCNN) to create trajectory classification model based on sequential patterns. The experimental results reveal that the proposed semantic analysis and text mining are not only dramatically suitable for indoor trajectory recognition, but also outperform other comparison methods on training error and testing precision.</p>