The establishment of an industrial benefit distribution mechanism for reservoir migrants in water conservancy and hydropower projects is a crucial part. This paper establishes the logistic regression model with gradient descent by deriving the objective function and using the gradient descent method to solve its approximate solution. Then, the model is applied to predict the total income and total expenditure of reservoir migrants before and after agricultural resettlement and tertiary resettlement, and the criteria for benefit compensation distribution are given based on the comparison before and after relocation. In the case of agricultural resettlement, the average share of equal compensation in the first 5 years accounts for 54.67% of the predicted income, and the predicted average annual growth rate of net income per capita is 17.86% with a discount rate of 3.12%. In the tertiary placement case, the predicted growth rates of wage income in the first 5 years are 7.49%, 4.84%, 10.56%, and 1.15%, respectively, and the average annual growth rate of transfer income is 7.01%, and the average annual growth rate of government subsidies is 0.80%. The prediction accuracy of the logistic regression model based on gradient decline reached 84.14%, and the analysis of the distribution of industrial benefits for reservoir migrants was credible. To let the living standard of migrants recover and exceed the level of non-relocation as soon as possible, it is necessary to give migrants certain compensation for production and life recovery based on the living standard measurement index.