In this study, the classic Boston house price data set is selected for the analysis of house price correlation. According to the variables in the Boston housing price data set, the linear regression model of Boston housing prices is established by using Python software. The regression equation and regression coefficient were tested for significance, excluding the variable of p >=0.5, multiple linear regression was carried out, and the regression equation with good fitting was obtained. It is found that there are too many variables after multiple linear regression, which is difficult to analyze and predict, so the correlation analysis of variables is carried out. This paper gets the percentage of lower status population, pupil-teacher ratio and average number of rooms per dwelling with medv (The median quoted price for an owner-occupied home, $1,000 per unit) has a significant relationship. Finally, a linear regression equation is established for the independent variable whose correlation coefficient is greater than 0.5, and the housing price is predicted.