Studies On The Forecasting Of The Main Disease In Cage-cultured Pseudosciaena Crocea

Along with the enlargement of farming scale and the degradation of traits, the frequency and the scale of occurrence of disease of Pseudosciaena crocea in marine cages were growing more serious, causing serious economic losses. In this paper, according to the role of the environmental factors (EV) and the climatic factors (CV) on disease of Pseudosciaena crocea in marine cages, the multiple linear regression, gray theory and BP neural network were applied to forecast main diseases. The main conclusions are as follows:1 The cage culture of Pseudosciaena crocea caused a certain influence on water quality. The chemical and physical index, such as the contents of suspended matter, phosphate, nitrate nitrogen, inorganic nitrogen, COD, DO, pH and chlorophyll in cage-cultured areas were significantly different. It also caused the decline of biodiversity, the change in community structure and the increase of dominance. The effect of environmental factors caused by cage culture was directly related to the incidence of cultured Pseudosciaena crocea.2 Using the surveyed data of the disease of Pseudosciaena crocea and EV and CV in Zhoushan from 2002 we analyzed the main disease and the effect of EV and CV to morbidity of main disease with the correlation analysis and the grey correlation analysis. The result showed that the main disease in cage-cultured Pseudosciaena crocea was bacterial disease which constitutes 88.9% percent of total incidence. And the significant factors were water temperature, transparency, suspended matter, COD, temperature, wind power, wind direction and bacteria content.3 Using the stepwise regression analysis, set up a multi-linear regression model to forecast the main disease of cultivated Pseudosciaena crocea. The forecasting model showed an high simulation effect.4 Using the grey analysis results, try to set up series grey model, including the GM(1,1), GM(1,2), GM~1(1,2), GM~2(1,2). The average forecasting precisions was 71.35%.5 Combining the correlation analysis results and the theory of BP neural network, we used the BP neural network to forecast the main disease of cultivated Pseudosciaena crocea, the forecasting results showed a high simulation effect, and the average forecasting precisions was 81.53%, which were satisfactory.6 Comparing the three kinds of model, the multivariate linear regression model can be used in the prediction of linear system, but it needs a large sample size and requires a certain probability distribution; Grey model can be used for "small sample", "poor information", overcoming the regression model of sample size and probability distributions requirement deficiencies, but because of its itself, this kind of precision restrictions model is more suitable for short-term forecasting; The BP neural network predictive model has a strong nonlinear mapping capability, which is suitable for solving the internal mechanism of complex problems.