Application Research Of Fishing Ground Forecasting Based On LightGBM And LGB-NN Models

In China,marine fishing industry is an important component in Marine fishery.Compared with developed fishing countries,Chinese fishing industry has many problems such as backward fishing equipment,uneven distribution of fishing areas and lack of scientific research.Therefore,improving the level of Chinese marine fishing industry is the key in development of Chinese marine fishery.In recent years,due to overfishing and environmental pollution,resources of Chinese offshore fishery are gradually declining,which leads to the decrease of temporal and spatial distribution stability of offshore fishing grounds.This paper aims to provide accurate information of forecasting fishing ground.Using fishery data and meteorological data,this paper proposes a fishing ground forecasting method based on Light GBM model,and proves the effectiveness of Light GBM model through experiments.On this basis,the Light GBM model is improved,and a fishing ground forecasting method based on LGBNN model is proposed.The improvement effect is verified by relevant experiments.The main work of this paper includes the following aspects:Firstly,this paper summarizes research progress related to fishery forecast at home and abroad in recent ten years,introduces the basic theories,modeling ideas and common prediction models of forecasting fishing ground,and describes the basic principles of Light GBM,DNN and GBDT+LR models in detail.Secondly,this paper collects fishery data and meteorological data related to forecasting fishing ground,and conducts data processing and feature engineering on these data,and constructs datasets of forecasting fishing ground.Thirdly,a fishing ground forecasting based on Light GBM model is designed and implemented.Firstly,this paper describes the construction of Light GBM model,model parameter setting and other detailed information.Then using different datasets,this paper conducts comparison experiment of forecasting fishing ground between Light GBM model and GBDT,XGboost models.The experimental results show that the forecasting method of fishing ground based on Light GBM model can get better prediction results.Fourth,based on Light GBM model,LGB-NN model is proposed and applied to fishery forecast.Firstly,this paper describes the construction of LGB-NN model,model parameter setting and other detailed information Then the prediction results on different datasets are compared with the fishing ground forecasting method based on Light GBM model.The experimental results show that the prediction effect of LGB-NN model is better than Light GBM model.