Abstract.Based on machine learning technique and oil well efficiency project practical problem, to the complicated circumstance of oil well efficiency, non-linear machine learning support vector machines ( SVM ) shows a better analysis results than the classified prediction result of linear machine learning logistics regression ( LR ). This paper analyzed and derived the theorems and classification reason of logistics regression and support vector machines. The experiments calculated and compared the accuracies of these two algorithms under the same conditions, the result conforms the conclusion.