Agricultural Soil Data Analysis Using Spatial Clustering Data Mining Techniques

As an unsupervised learning method, spatial clustering has emerged to be one of the most important techniques in the field of agriculture for soil data analysis. Soil data analysis is usually related to practice in agricultural production management or discovery in agro-ecosystem process, so it is not easy to obtain labeled data that requires human intervention, and it is also not realistic to set specified pattern in advance. It is desirable to review the research work on soil data analysis using spatial clustering techniques in context of agricultural applications, which is the object of this survey. Soil properties (including physical, chemical, and biological properties) and the characteristics of the spatial soil data are first introduced. Spatial clustering techniques are then summarized in five different categories. Soil data analysis using spatial clustering is reviewed in four categories of agricultural applications: agricultural production management zoning, comprehensive assessment of soil and land, soil and land classification, and correlation study for agro-ecosystem. The traditional clustering algorithms generally work well, and prototype-based clustering methods are more preferred in practice. Some machine learning models can be further introduced into the spatial clustering algorithms for better accommodation to various characteristics of soil dataset.