Planting Suitability Analysis Of The Tropical Fruit Tress Based On ALS-DBN

In recent years,the living standards of our people have been continuously improved.The demand for tropical fruits including dragon fruit,wax apple and mango is also gradually increasing,which makes the industrial scale of our tropical fruit grow rapidly.However,the frequent extreme weather in recent years has caused great losses to tropical fruit production.Therefore,the research on the planting suitability of tropical fruit trees in Guangxi region is of great significance not only reduce the disaster losses of fruit farmers and enterprises,but also to optimize the layout and development of tropical fruit industry in Guangxi in the future.Through literature research,it is found that the current domestic fruit tree and crop planting suitability analysis mainly uses the mathematical statistics method to zoning the planting suitability.However,the relevant environmental information affecting tropical fruit tree planting is complex and numerous,and the traditional mathematical statistics methods are difficult to express the intrinsic characteristics of complex information and with low adaptive ability.Therefore,in this paper,it is studied the analysis model for suitability of tropical fruit tree planting based on Deep Belief Network.Through in-depth study of the Deep Belief Network(DBN)model,it is found that if the learning step size is improperly selected during the training process of the DBN model,it is easy to cause problems such as slow training speed or gradient oscillation.A new Adaptive Learning Step(ALS)-DBN deep learning model is proposed for this problem.The ALS-DBN IImodel introduces adaptive learning step size algorithm and momentum term in the original DBN model.At the same time,the reliability of the ALS-DBN model is verified by using dataset of the KEEL database and dataset of the UCI database.The experimental results show that the ALS-DBN model has excellent performance in convergence speed and classification accuracy.Finally,an ALS-DBN-based tropical fruit tree planting suitability analysis model was built on the PyCharm platform.The influence of training parameters on the model in the ALS-DBN model was analyzed experimentally,and the optimal neural network structure of the ALS-DBN model was determined.Based on more than 4,000 actual meteorological and geographic information data collected by 50 meteorological observatories distributed in various cities and counties in Guangxi,and based on previous research experience and discussion results with relevant experts,12 environmental and climatic factors that have influence on suitability analysis of tropical fruit trees were selected.And simulation experiments were carried out based on the proposed model.The experimental results show that the ALS-DBN model has excellent performance in planting suitability analysis of dragon fruit,wax apple and mango.