基于改进AdaBoost算法对环柄菇毒性判别研究

Study on the toxicity identification of Pleurotus ostreatus based on improved AdaBoost algorithm

使用改进的AdaBoost算法,根据环柄菇的特征对其毒性进行判别,提高环柄菇毒性的判别速度和准确性。首先通过计算数据集的初始权值分布,根据权值分布对训练集进行学习,得到弱分类器,根据弱分类器的分类结果,将错误的样本权值加大,同时计算出各个样本特征的权值,并删去权值系数小于规定值的特征;计算每个弱分类器在训练集上的误差率,并计算该弱分类器在强分类器中所占的权重,更新训练集中的权值分布,对权值过高的样本加上了损失函数,得到了最终的强分类器,输出预测结果;最后通过混淆矩阵,准确率,支持率等指标定量计算改进前后的AdaBoost算法的优化程度。在试验证明部分,使用python对数据集进行清洗,并归一化处理,然后分别利用三种算法对数据集进行预测,得到各自的准确率,改进之后的AdaBoost算法准确率达到99.96%,比单一的弱分类器和改进前的AdaBoost算法准确率平均提高7.5%。该方法可以提升环柄菇毒性判别的准确性,缩短毒性判别的检测时间,推动食用菌毒性诊断专家系统的研究。

The improved AdaBoost algorithm is used to determine the toxicity of Pleurotus ostreatus based on the characteristics of Pleurotus ostreatus,and to improve the speed and accuracy of determining the toxicity of Pleurotus ostreatus.First,by calculating the initial weight distribution of the data set,the training set is learned according to the weight distribution,and the weak classifier is obtained.According to the classification result of the weak classifier,the wrong sample weight is increased,and the characteristics of each sample are calculated.Weight,and delete the features whose weight coefficient is less than the specified value;calculate the error rate of each weak classifier on the training set,calculate the weight of the weak classifier in the strong classifier,and update the weight in the training set Value distribution,the loss function is added to the samples with too high weight,and the final strong classifier is obtained,and the prediction result is output;finally,the confusion matrix,accuracy rate,support rate and other indicators are used to quantitatively calculate the optimization degree of the AdaBoost algorithm before and after the improvement.In the test proof part,use python to clean the data set and normalize it,and then use three algorithms to predict the data set to get their respective accuracy rates.The improved AdaBoost algorithm has an accuracy rate of 99.96%,which is better than single the accuracy of the weak classifier and the improved AdaBoost algorithm is improved by an average of 7.5%.This method can improve the accuracy of the toxicity identification of Pleurotus ostreatus,shorten the detection time of the toxicity identification,and promote the research of the edible fungus toxicity diagnosis expert system.