Research Of Key Technology Of AFC Dynamic Data Warehouse Application System

In order to meet the needs of subway operators and managers of value mining for the data stored in the system,part of cities introduce data warehouse into their subway system,but there are just historical data in the data warehouse,which can not provide powerful data support for the strategic and tactical decision-making in the operation and management of the subway.Therefore,dynamic data warehouse system is proposed to introduce into the subway system by this paper,build the AFC dynamic data warehouse application system,and do research on the key technical problems needed to solve in the AFC dynamic data warehouse application system construction process,namely the change data capture,the data cleaning and query contention issue in the process of data store and access.Firstly,the research status of change data capture technology,data validity detection and data-store-access-process technology in dynamic data warehouse are summarized by this paper,and then the data warehouse and dynamic data warehouse are introduced in detail.The technical principle and advantages and disadvantages of the existing main change data capture methods are analyzed in detail,and the performance requirements of the change data capture module in the AFC dynamic data warehouse application system are analyzed in detail.Finally,the log-based method is used to capture the change data in the ACC system.At the same time,the problem of log reading and analysis in this method is discussed,and the corresponding solution is given.Then,based on the metro data of the Nanjing subway,the abnormal detection and processing methods of the traffic data are studied.Because the anomaly detection threshold range computation is based on the forecasting model of the inbound passenger flow,so construct the forecasting model of the inbound passenger flow can be the first need to do.The chaotic characteristics,Time delay and optimal embedding dimension of the inbound passenger flow time series data are verified and determined by the C_C method and the Wolf method,and then the phase space reconstruction method is used to reconstruct the inbound passenger flow time series data to obtain the model training sample set,and the training sample set data is normalized later.Based on the vector regression model,the integrated learning method is used to construct the forecasting model of the inbound passenger flow.Based on the idea of Bagging and RF(Random Forest),the sub-models in the model integration are generated,and the sub-models are intergrated by model and then come out the intergrated model,and the particle swarm algorithm is improved by adding artificial interference particles(recorded as WPSOwGS),and the WPSOwGS is used to optimize the parameters of the integrated model.The data of the Daxinggong Station and Shanghailu Station are regarded as example data,the model construction and model parameters optimization are carried out according to the method described in this paper.At the same time,the threshold range calculation and data validity detection and processing are carried out.The experimental results show that the WPSOwGS overcomes the problem of easy to fall into the minimum point and has a relatively good parameter optimization effect.At the same time,the prediction model constructed according to the method has obvious advantages in the model prediction effect compared with the reference method.Compared with the reference method,the false detection rate of anomaly detection is obviously reduced when the anomaly detection threshold range is narrowed obviously,namely the anomaly detection and processing method has superior data detection and process performance.Then,The solutions of the existing query contention problems are comprehensively analyzed and discussed in this paper,and combines with the specific application requirements of the system,finally the dynamic multi-level cache method is chosen to deal with the query contention issue,and then according to the conclusion has been completed,the AFC dynamic data warehouse application system's architecture design work is accomplished.In general,some useful explorations on the problem of change data capture,data anomaly detection and processing and the query contention issue in data-store-access process in the construction process of the AFC dynamic data warehouse application system are made in this paper,which provides theoretical and methodological support for the practical construction of AFC dynamic data warehouse application system.