| In the process of cultivated land quality data investigation and collection, there will be data missing due to human, environmental and other factors. However, the current data missing filling methods have insufficient applicability. In order to improve the cultivated land quality database and improve the accuracy of cultivated land quality evaluation, it is very important to study the missing data filling methods of cultivated land quality evaluation. In this paper, the cultivated land quality database of Conghua district is taken as the sample set. According to the spatial correlation and spatial distribution, the data set is divided into spatial correlation and non spatial correlation data sets. A variety of filling methods are used to simulate the missing filling, and the cross method is used to verify the accuracy. The results show that: 1) the proportion of total outliers is less than 1.2%, and 25 groups of elements such as elevation, temperature and available zinc have spatial correlation. 2) The four image nearest neighbor algorithm has the highest filling accuracy for spatial association data, and the accuracy is as high as 80% when the missing rate is less than 20%. The accuracy decreases with the increase of the missing rate, followed by K nearest neighbor algorithm, expectation maximization algorithm, multiple filling algorithm and regression model algorithm. The four image nearest neighbor algorithm has better accuracy for k nearest neighbor algorithm when the data is dense. 3) For non spatial correlation data, the highest filling accuracy is similar aggregation filling algorithm, which can keep more than 80% accuracy within 25% of the missing rate, followed by expectation maximization method, multiple filling method and regression model algorithm. To sum up, the two filling methods proposed in this paper show higher accuracy, more stable effect and wider practicability than other algorithms in filling missing data of cultivated land quality evaluation. |
