Research on Embedding Unsupervised Learning and Application on Tobacco Leaf Data

Data-driven is the core of distance measurement, which is used to measure and classify the similarity of tobacco leaf producing areas and improve the accuracy of clustering. The article proposes a spectral clustering algorithm based on auto encoding called AESC, which consists of two parts: auto encoder and spectral clustering layer. The auto encoder is used for feature learning to improve the accuracy of distance construction, and the spectral clustering layer makes the algorithm more adaptable to data distribution. After experiments, the autoencoding spectral clustering algorithm proposed in this paper is higher than the traditional Kmeans, DBscan and spectral clustering algorithms in the Silhouette Coefficient. And the similarity measurement results are mutually verified with research in the tobacco industry, which proves the reliability of the algorithm for the similarity measurement of tobacco leaf quality and is of great significance for improving the stability of cigarette quality.