The returns of wealth management products are influenced by multiple factors such as the international situation, and the prices exhibit high noise and non-linear temporal data characteristics, making forecasting difficult. To optimize the performance of forecasting the returns of bank wealth management products driven by financial technology, this paper proposes a Faster R-CNN-BiLSTM-Transformer wealth management return forecasting model by means of multidimensional features driven by financial technology. Firstly, extract multidimensional femployed text semantic features of financial products and users, and the syntax and semantic information in the names of financial products are captured through convolutional neural networks (CNN); Then, a BiLSTM Transformer by means of gated recurrent units is introduced to optimize the temporal feature extraction mechanism, and a user feature forecasting model is constructed by combining additive attention module and multi head self-attention module; Finally, a forecasting ranking forecasting model for candidate financial products is constructed, which calculates the similarity between users and financial products to find the most suitable financial product for user needs. The results suggest that the multidimensional features Faster R-CNN-BiLSTM-Transformer framework proposed in this paper outperforms the CART regression tree forecasting model and ARMA statistical forecasting model in all evaluation indexes, significantly improving the performance of forecasting the index return rate of financial products.
