摘要
设计了一种多传感器与机器学习相结合的甲醛检测系统。通过多个传感器采集环境中的温湿度、甲醛、酒精、氨气的值。将采集到的数据利用局部异常因子(LOF)算法进行预处理去除异常值,然后通过决策树(CART)算法提取重要特征,再利用多项式核函数将重要特征映射到高维得到最终的特征向量,最后通过XGBoost算法对特征向量进行训练生成预测模型,将训练好的模型导入手机后进行推理,得到抑制交叉干扰后的甲醛检测结果并显示。
In the paper,a formaldehyde detection system combining multi-sensor and machine learning is designed.The temperature,humidity,formaldehyde,alcohol and ammonia values in the environment are collected by the mul-sensor.The collected data is preprocessed by the local anomaly factor(LOF) algorithm to remove the outliers,and then the important features are extracted by the decision tree(CART) algorithm.Then the polynomial kernel function is used to map the important features to the high dimension to obtain the final feature vector.Finally,the XGBoost algorithm is used to train the feature vector to generate the prediction model.After the trained model is imported into the mobile phone,the prediction is performed,and the formaldehyde detection result after suppressing the cross interference is obtained.
作者
邹卓娟
陈向东
Zou Zhuojuan;Chen Xiangdong(College of Information Science and Technology,Southwest Jiaotong University,Chengdu 610031,China)
出处
《单片机与嵌入式系统应用》
2019年第7期23-26,55,共5页
Microcontrollers & Embedded Systems