This paper combines the cigarette quality supervision measurement project, summarizes the key instruments for cigarette quality supervision, constructs the reliability evaluation model of cigarette quality supervision measurement instruments, adopts the random forest method and carries out model optimization so as to solve the weights of quality evaluation indexes of cigarette quality supervision measurement instruments. The QTM comprehensive test bench is selected to measure the quality and circumference of the traceability comparison stick of cigarette physical inspection equipment and combined with the Random Forest optimization algorithm to analyze the impact of measurement repeatability on the moisture content of cigarettes. Using the normalized deviation evaluation method, we can compare the consistency of the detection results of the evaluation model of this paper with the detection results of the parent company. It is analyzed that the maximum permissible error of cigarette circumference is ±0.05, and the maximum permissible deviation of comparison Dmax is ±0.06. The circumference deviation of sample A is 0.04<0.06, the sample is qualified, and it is consistent with the testing results of the parent company.