The invention discloses a dissolved oxygen proportionality improvement method. The method comprises the following steps: 1, generating a submerged jet flow in a horizontal direction by a submersible pump; 2, repeatedly carrying out flow jetting experiments in a water area, using an extreme learning machine to carry out modeling to data obtained by the flow jetting experiments, obtaining relations between a dissolved oxygen discrete coefficient and an initial dissolved oxygen discrete coefficient, time and water pump parameters, wherein the network structure is an ELM neural network structure; 3, measuring the initial dissolved oxygen discrete coefficient before some flow jetting; 4, placing the initial dissolved oxygen discrete coefficient in a model to determine plug flow time; and 5, carrying out flow jetting to the water area, and ending the flow jetting when the flow jetting time reaches the time determined in the step 4. The method utilizes the submersible pump for flow jetting to improve the dissolved oxygen proportionality, utilizes an artificial neural network to establish the relation between the flow jetting and the dissolved oxygen equalization index, and utilizes the relation to predict flow jetting time and guide flow jetting.