Research On Location Recommendation Of Urban Air Quality Monitoring Equipment Based On Association Rules

The distribution of air quality in urban areas is closely related to people’s physical and mental health.If the air quality distribution can be known in real time,it will have a pretty important reference significance for the prevention of residents’ outdoor activities and the subsequent environmental governance of relevant government departments.The distribution of air pollutants has spatio-temporal interactions,and is affected by many complex external factors,such as weather,traffic volume,road network structure,and land use.In addition,the area covered by established air quality monitoring stations in urban areas is relatively sparse compared to the entire urban area.These series of factors make it challenging to predict the fine-grained air quality distribution in urban areas.However,the existing air quality prediction methods either do not take into account the influence of external factors related to the air quality distribution,or it is difficult to capture the spatio-temporal interactions of the air quality distribution due to the limitation of its own model performance.On the other hand,with the continuous expansion of the urban area,the air quality monitoring network originally established in some cities can no longer meet the accuracy requirements of the entire urban area air quality monitoring.Based on the established air quality monitoring stations in the urban area,how to recommend a limited number of optimal station locations from the candidate areas,so as to build a more effective and purposeful air quality monitoring network with the original monitoring equipment,which has become an urgent problem for air quality management in many cities in our country.In this way,the new monitoring network can timely and accurately reflect the distribution of air quality and its development trend in the entire urban area.However,most of the current research on this issue cannot be based on the established air quality monitoring stations,the general flexibility is relatively low.In addition,the monitoring items are mainly conventional pollutants,which cannot meet the monitoring requirements of combined air pollution caused by coal smoke and motor vehicle exhaust.In response to the above-mentioned problems and challenges,this paper considers the external influential factors related to air quality distribution,and divides the target city into disjoint grids.Each grid is a sub-area of 1km*lkm and is associated with time-varying air quality values.We take each grid as a node and construct a urban spatio-temporal graph.Then we carried out a research on location recommendation method for urban air quality monitoring stations based on association rules.The main research contents are as follows:(1)Aiming at the prediction problem of air quality distribution in urban areas not covered by monitoring stations,we formulate the spatial fine-grained air quality prediction problem as a spatio-temporal graph prediction problem.We have designed a spatial fine-grained air quality inference model based on high-order graph convolutional network(HGCNInf).HGCNInf uses historical air quality data monitored by sparse monitoring stations to capture the spatio-temporal interactions of air pollutant distribution,combined with external factors such as urban road network structure,urban weather,and land use POI,so as to effectively predict the spatial fine-grained air quality distribution in urban areas not covered by air quality monitoring stations.(2)Aiming at the location recommendation problem of the air quality monitoring station,we formulate the monitoring station location as a urban spatio-temporal graph node recommendation problem in which each node represents a region with timevarying air quality index.Based on the air quality inference model HGCNInf,we have designed a greedy algorithm for minimizing information entropy(GMIE),which aims to mark the recommendation priority of candidate areas according to the ability to improve the inference accuracy of HGCNInf through the node incremental learning method.Finally,we recommend the node with the highest priority as the new monitoring station location,which could bring the greatest accuracy improvement to HGCNInf.We evaluate the proposed model by using real-time air quality data of the target city and experimental results show that our approach far outperforms state-of-the-art baseline methods.