TAR: A Highly Accurate Machine-Learning Model to Predict the Cocoon Shell Weight of Tasar Silkworm Antheraea Mylitta
摘要
In this paper, we propose a machine-learning model for predicting the shell weight of silkworm cocoons Antheraea mylitta D. (Saturnidae) without cutting open the cocoon. Our proposed work uses a topology adaptive kernel regression (TAR) to predict the shell weight of cocoons based on a set of non-invasive easy-to-measure cocoon features. We evaluate our model on four datasets from different families of cocoons. The evaluation shows that the proposed model accurately predicts the shell weight and outperforms well-known models, including neural network-based regression.