At the final stage of oil field development, the problem of maintaining acceptable levels of oil production through operational management of flooding is an urgent one. The complexity increases due to the growing number of wells in the basin and the variability of their interaction. This requires new approaches that take into account these trends in oil production. Two algorithms focused on large dimensions and small data samples are proposed. They were tested on model data, in which 60 injection wells, 160 production wells and 17 observations, random errors were present. These algorithms have demonstrated acceptable performance both in terms of accuracy and speed, and in terms of the possibility of using them for predicting oil production.