The invention discloses an ultrasound breast tumor grading method based on multi-feature extraction and Linear SVM. The method comprises steps that to-be-diagnosed breast ultrasound RF data is inputted; the RF data is converted into an ultrasound B-mode image to obtain the initial suspicious lump area location; the initial suspicious lump area is segmented to obtain a suspicious lump to determinethe boundary of the suspicious lump; a feature value of the segmented suspicious lump is calculated; the calculated feature value is inputted into a classifier, and the suspicious lump area is analyzed. The method is advantaged in that based on the BI-RADS grading standard, the actually acquired breast ultrasound RF data is hierarchically detected, breast tumor ultrasound 3, 4 and 5-level classification is achieved through utilizing the multi-feature extraction method and the linear support vector machine classifier, improvement of accuracy of the doctor's diagnosis is facilitated, the algorithm time and complexity can meet clinical requirements, and the method has a certain practical value.