Water Quality Abnormal Pattern Recognition For Real-time Monitoring Of Water Pollution

In recent years,a large number of water pollution incidents have occurred in China,threatening the drinking water safety of the people and destroying the water resources ecosystem,causing incalculable losses.How to effectively establish a water quality monitoring system,provide timely warning of water pollution,and prevent it from becoming a problem that needs to be solved urgently in the field of water environmental protection.Abnormal water quality data refers to anomalies that cause water pollution environmental monitoring data of water resources,timely and accurately discover the abnormal state of water quality in water resources environment,explore the relationship and laws between water quality occurrence environment and state,and preventive protection of water resources.Providing a scientific basis is a challenge for current water resources warning.In a big data environment,people are more concerned than ever about ways to quickly get valuable information from all data.Based on the research and analysis of the advantages and disadvantages of the previous unsupervised anomaly detection algorithm based on the real-time monitoring of the water quality monitoring data,this paper proposes an unsupervised anomaly detection algorithm based on unlabeled water quality.The data set discovers the abnormal state in the water quality in real time to help the management department make timely decisions.The main research content of this paper includes the following three parts:1.Select the k-means clustering in the clustering algorithm to cluster the water quality data.The water quality anomaly data can be detected by clustering,and the data can be reduced to 3D by PCA principal component analysis,and then the visual clustering effect is performed.To evaluate the clustering effect,the contouring coefficient can also be used to quantify the clustering effect of the data.2.The clustering result of the previous part is used as the true value.Firstly,the clustering result is randomly forested to obtain the dimension importance,and the water quality data corresponding to the anomaly can be obtained.Which features are most important for the detection of abnormal water quality data,That is to say which elements have the greatest impact on water quality anomalies,and this conclusion is of great significance for the prevention and classification of water quality anomalies.3.Through the above research,establish a suitable supervised classification model.After similar data,you can directly classify the model and get whether the data is abnormal water quality data.Comparing the accuracy of various basic classification algorithms and integrated learning classification algorithms,it is concluded that the integrated learning classification algorithm is better than the basic classification algorithm,and the best effect is GBDT algorithm.