In order to improve the classification performance of financial text information and reduce the loss value of financial text information classification, this paper proposes a financial text information classification method based on attention mechanism and multi-dimensional feature fusion. Using a vector space model to represent text, establishing an appropriate feature evaluation function, and selecting feature items. Input the processed financial data into the PCNN neural model for denoising processing; Capture the correlation information between long-distance contexts in text and extract temporal information from text sequences using temporal convolution; Through the channel attention mechanism SENet, the deficiency of losing contextual information in the local receptive field in convolutional computation is compensated for; Use BERT pre trained model to reinforce sentence semantic features with raw dynamic word vectors. Calculate the probability distribution of the category, select the category label corresponding to the highest probability as the prediction result, and output the information classification result. The results show that the convergence speed of the model in this article is fast, and the loss value of the model is the lowest, indicating that the classification results of the model using the classification method in this article have the smallest error compared to the actual results, and the predicted labels are closest to the true labels. In practical applications, obtaining more accurate analysis of financial data.