In order to accurately understand the economic development of enterprises and increase the company’s economic benefits, a study on financial forecasting and decision-making in big data cloud accounting enterprises is proposed. Enterprises improve the efficiency of data utilization by acquiring information processing and analysis, establishing a diversified control mechanism, and improving the effectiveness of financial and tax management. The objective function is optimized using a structured sparse induced parametric number to calculate the data block centers to describe the data objects more comprehensively and make the obtained clustered financial results more accurate. Adding classifiers to the set of labeled samples and constraining the joined samples belonging to the wrong class combine multiple kernels from different perspectives to obtain a comprehensive measure of similarity. Selecting sub-kernel functions and parameters to construct multiple kernel functions, the learning and generalization capabilities of kernel functions, and using high-dimensional data feature vectors to construct a shared hidden subspace to maximize the similarity between prediction samples and assign greater weights in the multi-perspective clustering process for corporate financial forecasting and decision making. The analysis results show that using data clustering cloud finance, financial data can be collected and corrected promptly, and the budget accuracy is up to 90%, which provides important help to enterprise financial decision-making.