Weight Analysis Of Each Influence Factor During The Green Tide Disaster Based On SVM

Green tide refers to the phenomenon that Enteromorpha floating and widely propagated in ocean, which has caused varying degrees of economic losses and environmental influences. Therefore, the causes of green tide has become the focus of attention. As the observation data of green tide is precious and rare, what’s more, the cause of the green tide disaster is extremely complex, which would cause some difficulties to the factor analysis and disaster predicting during the development of green tide.In order to predict the disaster better, this thesis studies the factor analysis of the green tide disaster. According to the continuous observation data of factor, which influences the growing and spreading of green tide, and the interrupted yellow green tide distribution area（DA）, the thesis analyzed the development process of green tide disaster and proposed the factor’s weight analysis algorithm based on SVM. Then through the analysis for the change rule of weight, give the division basis of the various stages of green tide. The proposed methodology provides a new way for the factors’ analysis of the other marine disaster. The contributions of the thesis work can be summarized from the following four aspects.First, aiming at the problem that the factors causing green tide disaster are extremely complex, the thesis analyzed the mechanism of bloom and then summarizes its origin, development process and affect factors. And concluded the major causes of green tide including temperature（T）, weather phenomenon（WP）, wind direction（WD）,wind force（WF） and wave height（WH）, which establish the foundation for the sample the data constructing of the model.Second, in view of the problem of small observational sample size and the causesof disaster with high complexity, combining the SVM’s advantages that can map the data from observation space to feature space to learn better, in order to restore the continuity of the observed data, the green tide DA interpolation model based on SVM is build. Considering the value of green tide DA in different periods is difference obviously, the nearest interpolation model based on SVM was put forward. The specific unknown data was predicted by the corresponding interpolation model and the continuous green tide observation samples were got finally. The green tide DA in2012²015 were interpolated by using the proposed algorithm and the recovered result accord with the change rules of DA during the process of green tide disasters, which shows that the algorithm is robust.Third, aiming at the problem that the influence factors will change in the different period of green tide disaster, the factor weight analysis algorithm based on SVM was proposed. Then analyze the influence weights of various factors on the green tide DA using this algorithm and got W_T、W_WP、W_WD、W_WF and W_WH. Following conclusions were obtained by analyzing and comparing each factor weight:（1） T is the primary influence factor that affects the green tide disaster.（2） The main force of the movement of the green tide in ocean is WF and the sustained strong wind is the main external force making the green tide expand gradually.（3） WP have no significant effect on green tide in the outbreak stage.（4） WH affects the satellite remote sensing monitoring of DA in the disappear stage. The analysis of temperature weight difference（DW_T） shows that: the DA starts to get bigger when DW_T hops, green tide enters into the outbreak stage while DW_T reaches a maximum, with the decrease of DW_T, green tide steps into the stage of dying slowly. Therefore, the time dividing points of green tide’s float stage, outbreak stage and dying stage can be determined according to the change of DW_T.Finally, the disaster situation of green tide in 2014²015 was predicted using factor’s weight analysis algorithm based on SVM. The relation compared between green tide DA and DW_T shows that DW_T present consistent change rules during each periods of the year, which shows the algorithm has strong robustness.