Text data-based model for analysis of floating population residence analysis

Floating population management is one of the most fundamental parts of the grassroots policing. Traditionally, a practice of door-to-door visits and manual analysis is used in all the regions in China to collect floating population residence information but it is time consuming and inefficient due to large data size. In view of this, a floating population residence analysis model is proposed in this study in which floating population text data is collected and processed by the Word2vec weighted word vector model to vectorize the text data. The vectorized data were then put into the convolutional neural networks (CNN) for learning and training, for faster recognition and classification in the floating population registration management. The model can be used with 81%, prediction accuracy and can substantially improve the efficiency of the grassroots public security organs tackling the challenges in the existing manual analysis of floating population residence.