Modern Economics & Management Forum

With the rapid development of information technology, data is being generated and accumulated at an unprecedented scale and speed. The sources of economic operation data have expanded from traditional statistical surveys to multiple channels such as Internet transactions, social media, satellite remote sensing, financial markets and industrial chain logistics. Traditional statistical analysis and economic forecasting methods have become difficult to cope with the massive, unstructured and dynamic characteristics of data. This article unfolds from four aspects: machine learning regression optimization, the fusion of time series statistics and dynamic Bayesian, multi-dimensional factor dimensionality reduction analysis, and the cross-application of statistical learning and text mining, aiming to reveal the transformation logic and practical significance of traditional statistical models in the big data environment. By constructing a data-driven prediction framework, a leap from parametric inference to adaptive learning is achieved, providing technical and systematic theoretical references for economic trend analysis and macro decision-making.