Extended Probabilistic Latent Semantic Analysis Model For Topics In Time-Stamped Images

Abstract

Artificial neural network is a powerful tool for modelling and prediction in various fields. Tobacco management specialists need simple and accurate techniques to evaluate tobacco growth and cultivation in the growth period. The objective in this study is to develop a back-propagation neural network (NN) model to accurately predict tobacco growth under various climatic and soil conditions in China. The proposed model used the field test data at 11 locations in main tobacco areas in China. As the output of the NN, the dry weight of tobacco plants was significantly influenced by dry weigh preceding stage, time for sampling, average temperature, relative humidity, soil organic matter and available P value, so these 6 environment factors were taken as the NN input. In the study, an optimal NN model was developed with 6 inputs, 1 output and 19 neurons in the hidden layer. With the same test data, the proposed NN model could perform better than the conventional regression model in terms of the relative error ranges and relative coefficient.