[Purpose/significance] As an important platform for people to express their views,micro-blog has gradually become a research hotspot of text sentiment analysis. This paper proposes a LSTM model based on attention mechanism,and takes micro-blog comments of Huawei P10 problem as the research objects to analyze the emotional tendency of network users,which further verifies the effectiveness of this model. [Method/process] The paper introduces the depth learning theory,and uses the LSTM model to carry out emotional analysis in order to better grasp the emotional information in the text and enhance the success rate of emotional classification. [Result/conclusion]The LSTM model is effective,which can be reflected in the accurate analysis of the emotional features of longer texts,especially for paragraph texts of micro-blogs. [Limitations]The proposed model cannot deal with Yan text,facial expressions,and other non-text information,as well as reflect the relationship among words.