Study on the Method of Cultivated Land Quality Evaluation Based on Machine Learning

With rapid economic development, acceleration of urbanization and population growth, it causes many resource issues including soil pollution, soil erosion and unreasonable cultivated land use. More seriously, both cultivated land quantity and quality are decreasing greatly faster. Besides, with respect to physical reduction in the amount of cultivated land, the hidden decline in cultivated land quality is far more harmful to food security, ecosystem protection, and economic sustainable development. In fact, the quality of cultivated land is determined by the characteristics of different kinds of factors and the influence on each other. Therefore, an objective and accurate method of cultivated land quality evaluation is necessary and beneficial.In this paper, we analyze the association relationships of thirteen evaluation factors in the national cultivated land quality evaluation system by using FP-growth algorithm. According to the correlation results, we exclude those evaluation factors with high association relationship so that we can accomplish dimension reduction and improve evaluation efficiency under the premise of ensuring the quality of evaluation. Based on training and testing of BP neural network, the grade models of cultivated land physical quality grade are established. The methods avoid the influence of artificial factors such as experts' scoring in the model to determine the weight of every factors and some other human factors, so that improve the objectivity of the grade of cultivated land quality. Finally, we choose Guangzhou as a study area, using its cultivated land quality data for dimension reduction experiments. After training the grade models with massive data, we obtain the results of cultivated land physical quality grade in Guangzhou. According to the experiments' results, the accuracy rate of the cultivated land quality evaluation in Guangzhou can get with almost no loss. It can also show that the evaluation model of cultivated land quality given in this paper can be used at the case of that some data are missing or abnormal, and meet the expected accuracy.