基于梯度提升树的消防员体征数据挖掘与火场行为识别模型构建

Model construction for physiological data mining and fire scene behavior recognition of firefighters based on gradient boosting tree

为了进一步分析利用消防员体征数据,为提高现场救援效率提供支撑,通过“智能夹克”获取3448.4万条物联网监测数据,运用机器学习中的梯度提升树算法展开计算机实验计算。研究结果表明:当决策树为100棵时,消防员火场中80个行为状态动作识别的准确率由决策树深度决定,当决策树深度为1时,训练集的识别准确率为72.37%、测试集的识别准确率为53.11%;当决策树深度为4时,准确率均能达到100%;但随着决策树深度的增加,运算的时间成本将大幅度增加,因此,在确定决策树深度时,需要综合考虑时间成本与准确率之间的影响关系。

In order to further analyze and utilize the physiological data of firefighters,and provide support for improving the efficiency of on-site rescue,34.484 million IoT monitoring data were obtained by the‘smart jacket’,and the gradient boosting tree algorithm in machine learning was used to carry out the computer experimental calculation.The results show that when the quantity of decision tree is 100,the recognition accuracy of 80 behavior state action of firefighters in fire scene is determined by the depth of decision tree.When the depth of decision tree is 1,the recognition accuracy of the training set is 72.37%,and the recognition accuracy of the test set is 53.11%.When the depth of decision tree is 4,all the accuracies can reach 100%.However,as the depth of decision tree increases,the time cost of calculation will increase significantly.Therefore,when determining the depth of decision tree,it is necessary to comprehensively consider the relationship between time cost and accuracy.