Application of Heterogeneous Data Analysis based on SEA Grid in User Investment Analysis

Abstract

To solve the problem of yield calculation in complex investment scenarios, a time-cost double-weighted rate of return calculation method based on SEA grid is proposed, and its effectiveness is verified by comparing with  traditional methods and researching the quantitative evaluation and analysis method of user investment based on structured data. To solve the above problems, a data-heterogeneous federated learning method based on user investment analysis FedPSG is proposed, which changes the data form transmitted from the client to the server from model parameters to model scores, and only a small number of clients need to upload model parameters to the server in each round of training, thereby reducing communication costs. At the same time, a model retraining strategy is proposed, which uses server data to train the global model for second iteration, and further improves the model performance by alleviating the impact of data heterogeneity on federated learning. The method of event dimension analysis of user investment is designed, and a credibility index is proposed to evaluate the analysis results. Experiments show that by combining event data, it can effectively provide users with event factors in the fluctuation of investment profit and loss, and help users better analyze their own investments.