Application Of Data Mining Technology In Evaluation Of Degraded Grassland In Sanjiangyuan Region

The policy of "giving priority to ecological protection" put forward by Qinghai Province has raised people’s awareness of ecological protection to an unprecedented height.However,in recent years,due to the influence of natural factors and human activities,the grassland vegetation in the Sanjiangyuan region has been seriously degraded.Reasonable prediction of grassland degradation is the premise of protecting the ecological environment in this area.However,there is a lack of data of degraded grassland in this region,the traditional manual measurement method is time-consuming and inefficient,and there are few prediction models for degraded grassland in Sanjiangyuan region.Therefore,based on the actual work of grassland workers’ evaluation of degraded grassland in Qinghai Province,this paper collected the degraded grassland data of alpine meadow in Sanjiangyuan region,analyzed and mined the grassland with data mining technology,and provided theoretical guidance for grassland workers’ intelligent evaluation of grassland.The main work of this paper includes the following contents:(1)Due to the incomplete classification of degraded grassland in the original alpine meadow data of the Three-River Headwaters,it cannot be directly classified and predicted.In this paper,the accuracy of clustering and semi-supervised clustering was compared.The experimental results showed that the accuracy of AMPCK-means algorithm of semi-supervised clustering algorithm was the highest.Therefore,the algorithm was used to label the degradation categories in the original data to form a complete data set,improving the accuracy and integrity of the classification data,which were divided into four degraded grassland categories.(2)Use some classical classification algorithms to model the classification prediction of the data marked in(1).The experimental results show that the accuracy of the data annotated by AMPCK-means algorithm in the Random Forest model,Native Bayes model,SVM model and Logistic Regression model is 96%,92%,95% and 88%,respectively,achieving good accuracy.(3)The association rule Apriori algorithm was used to analyze the relationship between different grassland types and grass species at four degraded grassland levels.The results showed that with the aggravation of degradation,the superior grass species on grassland gradually changed into weeds and poisonous grass.(4)In order to improve the prediction accuracy of alpine meadow degradation in Sanjiangyuan and shorten the testing time,this paper proposed a grassland degradation prediction model based on the fusion of parallel K-means and DNN.Specifically,in order to solve the problem of insufficient data,sufficient data are collected on the basis of the above data and divided into training set and test set.The k-means algorithm is improved in parallel.In the test stage,the parallel K-means method is used to cluster the data,and then the trained DNN model is used for prediction.The experimental results show that the average prediction accuracy of the parallel K-means-DNN model is more than 99%,and the AUC index is more than 95%,which is better than the accuracy of the previous classification model.At the same time,it proves that the model can be well suitable for the prediction of the degradation of alpine meadow in the Sanjiangyuan Region.Finally,a grassland degradation evaluation system was designed based on the parallel K-means-DNN prediction model.