Model Predictive Control (MPC) is a multivariable control and optimization technology widely applied in various industries. The paper offers an alternative approach, which employs "point" models developed at each time step. To create such models, a data mining algorithm is used, which addresses the associative database of inductive knowledge. This knowledge is obtained by extracting patterns from data through intelligent analysis. The associative search algorithm calculates the coefficients of the predictive model and the control actions simultaneously at each time step. The algorithm for calculating these coefficients and actions for one or more time steps in advance is proposed.