基于随机森林算法的耕地面积预测及影响因素重要性分析——以甘肃省庆阳市为例

Prediction of Cultivated Land Area and Importance of Influencing Factors Based on Random Forest Algorithm——A Case Study of Qingyang City,Gansu Province

[目的]分析耕地面积变化影响因素的重要性,以便科学预测耕地资源数量,为保护耕地资源服务。[方法]以属于黄土高原地区的甘肃省庆阳市为例,尝试采用随机森林算法构建耕地面积预测模型,与BP神经网络模型的预测结果进行对比,并对耕地面积变化影响因素重要性进行排序。[结果]随机森林算法预测结果的相对误差和均方根误差均小于BP神经网络的,预测精度高,结果稳定。它预测出2020,2025,2030年的耕地面积分别为4.515×10~5,4.513×10~5,4.512×10~5 hm^2,呈现减少的趋势;主要影响因素重要程度排序为:农业机械总动力>农业人口>地区生产总值>固定资产投资额。[结论]随机森林算法适合于耕地面积预测,且能够测度耕地面积变化影响因素的重要程度。

[Objective]To analyze the importance of the factors that influence the change of cultivated land area in order to predict the amount of cultivated land area resources,and to service the protection of cultivated land.[Methods]Taking Qingyang City of Gansu Province as a case study,the random forest algorithm was used to construct the prediction model of cultivated land area.The results were compared with those of BP neural network model,and the importance of the factors that influencing cultivated land area change was sorted.[Results]The relative error and root mean square error of the prediction results of the random forest algorithm were smaller than that of BP neural network,and the prediction accuracy was high and the results were stable.The cultivated land area in 2020,2025 and 2030 was predicted to be 4.515×10 5,4.513×10 5 and 4.512×10 5 hm 2,respectively,showing a decreasing trend.The importance of the main influencing factors was ranked as:agricultural machinery general dynamics>agricultural population>GDP>fixed asset investment.[Conclusion]The random forest algorithm is suitable for the prediction of cultivated land area and can measure the importance of factors that influence the change of cultivated land area.