The construction of intelligent financial management platform for the integration of industry and finance in universities based on SMOTE-RFM

Addressing the problem of sample bias and behavioral feature sparsification in multisource data of university finance in artificial intelligence management, this study constructs an intelligent financial management platform based on synthetic minority class oversampling and proximate-frequency-amount model for university industry-finance fusion, and optimizes the performance of dynamic industry-finance fusion data balancing and spatio-temporal feature extraction. The experimental results show that the transaction processing speed of the platform reaches 1.8 times of the comparison methods; the budget execution deviation rate is reduced to about 50% of the traditional model. The platform's memory consumption is only 51% of the benchmark level, and the floating-point computation is reduced to 30% of the original method. In addition, the platform excels in nonlinear feature modelling and high-dimensional data adaptability, and the differences in deviation rate and computational efficiency are statistically significant (p⪅0.01), providing feasible support for the engineering deployment of intelligent financial management. In summary, the platform proposed in the study can improve the dynamic monitoring of financial behaviors and intelligent decision-making, and provide a lightweight, interpretable and integrated solution for the digital transformation of business-finance integration in universities.