基于土壤数据广度与深度模型的作物推荐算法

针对现有数值型作物推荐算法忽略了文本域数据对于作物推荐的指导性意义，无法挖掘数值域数据与文本域数据之间的内在关联，导致推荐模型预测精度较低的问题，提出了一种基于土壤数据广度与深度模型的作物推荐算法。对残缺、重复、不平衡的土壤数据进行数据预处理，采用数值归一化和向量嵌入的方法融合数值域数据与文本域数据，然后使用广度与深度模型联合训练的方法挖掘其内在关联，改进多分类激活函数实现多分类。实验结果表明:该方法的预测精度优于现有数值型作物推荐算法。

Algorithm for Crop Recommendation Based on Wide and Deep Model of Soil Data

The existing numerical crop recommendation algorithm has the deficiencies that ignore the guiding significance of text domain data for crop recommendation， and cannot explore the internal relationship between numerical domain data and text domain data， which leads to the problem of low prediction accuracy of the recommendation model.This paper proposes a crop recommendation algorithm based on soil data breadth and depth model. Firstly， pre-process of the incomplete， repeated， and imbalanced soil data was done， and then numerical domain data and text domain data were fused by employing methods of numerical normalization and vector embedding. At last，internal correlation was mined by using joint training of breadth and depth models，and multi-category recommendation was realized by improving the multi-category activation function. Experimental results show that the prediction accuracy of this method is better than that of the existing numerical crop recommendation algorithms.