In the information age, massive data not only brings convenience to people's lives, but also brings great challenges. How to quickly and accurately find the demand data from the massive semi-structured data and quickly compare and analyze the economic development trend is particularly important. This paper mainly analyzes and studies the application of deep learning in the intelligent prediction and analysis of urban economic statistics, introduces the individual cognitive items in PSO((Particle Swarm Optimization) algorithm into BA(Bat algorithm), and constructs an intelligent prediction and analysis of urban economic statistics based on the improved BA-LSTM model. The results show that the improved BA-LSTM model has the best fitting effect and the smallest error, which can reach 2.399. The experimental results show that the improved BA-LSTM model can accurately reflect the changing law of GDP growth. Therefore, LSTM model has high application value in intelligent prediction and analysis of urban economic statistics.