Application Of Machine Learning In Data Mining Of Earthen Archaeological Sites Monitoring

The Outer Burial Pits of the Han Dynasty Yang Mausoleum is the first fully enclosed site museum in China and the first site museum in China to effectively combine the protection of cultural relics with the display of cultural relics.As the cultural relics in the museum are not separated from the surface of the soil,the environmental factors on the surface and underground inevitably cause damage to the cultural relics.A large number of Io T sensors are used to monitor and evaluate the quality of the surrounding environment in the Han Dynasty Yang Mausoleum,hoping to discover the rules that are conducive to the protection of cultural relics from the monitoring data.So far,although more than 7 million pieces of various monitoring data are accumulated,the mining and analysis of this type of data is not comprehensive.There are the following problems: 1)Temperature data is used directly to modeling.Due to the absence of temperature data,the temperature prediction model built is not accurate.2)Only the average of temperature data is used to analyze the evolution laws,the data utilization rate is low,and it is difficult to ensure the accuracy of the mining regulars.In order to solve the above problems,the following work is done in this paper:1.Aiming at the problem of large missing temperature data,A data reconstruction algorithm based on Fourier basis matrix is proposed in this paper.The algorithm uses Fourier basis matrix as a frequency analysis tool.The coefficients of the Fourier series are updated by the gradient descent method,so that the difference between the inverse transformed data and the available original data is the smallest at the corresponding data point.The effectiveness of the proposed method is verified by the temperature monitoring data of the Outer Burial Pits of the Han Dynasty Yang Mausoleum.Comparative experiments demonstrate that this algorithm can effectively restore the integrity of museum temperature data,which solves the problem of missing temperature data caused by monitoring sensors hardware or network failures,and provide a complete and accurate data source for subsequent temperature prediction models.2.To solve the problem of the low utilization rate of temperature data and the inaccurate temperature prediction model.The extreme learning machine(ELM)in the one of machine learning algorithms is used to construct a new temperature prediction model in this paper.In this algorithm,ELM is used to analyze the data systematically and temperature prediction model is established based on the features of lunar calendar and solar terms.Experiments show that compared with the Gregorian calendar,the lunar calendar and solar terms are more closely related to temperature.At the same time,the comparative experiments demonstrate that the lunar calendar has certain advantages over the solar calendar in natural data recording and processing.