Assessment For Soil Fertility Quality Based On Genetic Algorithm And Fuzzy Neural Network

Nowadays, a wealth of approaches can be used for quality evaluation for soil fertility, for instance, aggregative index number, analytic hierarchy process, fuzzy mathematics and grey system, etc. Typically, they need membership function construction and confirming the weight value of evaluation index precisely and objectively, on top of that, the design of membership function may be bias to some extent.It is well acknowledged that Fuzzy-neural network has several advantages such as powerful self-adaptation, excellent sense of organization and fault freedom, which proliferated in information processing, automatic control, pattern recognition, etc. Therefore, it may be utilized for quality evaluation for soil fertility.In the present study, we took Mingshan River for illustration, by using EXCEL,SPSS16.0 and MATLAB7.1, (i) characterized the soil fertility index in distinct utilization type, (ii) assessed quality classification of soil fertility in catchment basin with geno- fuzzy-neural network mould, (iii) evaluated the soil fertility by using comprehend fuzzy-assessment, (iiii) confirmed and compared the results of evaluation. The conclusion can be drawn as follow:(1) The soil quality in small watershed is significantly affected by the utilization type. Specifically, the content of organic matter and nitrogen is on high level, indicating that paddy field may gather nutrient ran off from small watershed. Furthermore, the land with less mankind activity has more utilance in soil organic matter, full- nitrogen content and valid kalium, which has higher soil fertility. More exactly, Artificial bamboo forest took advantage in soil fertility recovery, which can curtail erosion of soil, soil & water loss and improve soil quality; Soil in tea plantation has higher content of organic matter, total nitrogen and immediate effect phosphours than arid land significantly due to that the distribution of arid land is mainly in hillside fields, whose nutrient(e.g. organic matter, valid kalium) severely ran off because of low vegetational cover and localized rainfall.(2) The neural network established the nonlinear mapping relationship between soil fertility factor and soil fertility states according to the explanation of Mingshan River soil fertility evaluation and based on self-learning of training sample, which can also evaluate and predict the fertility of candidate soil. The results showed that the paddy field has the highest quality classification of soil fertility, whose quality aggregative index number is 3.7411; the lowest one is the arid land. In terms of utilization type, the aggregative index number of soil fertility of each can be described by following order: paddy field> nature wood land> tea plantation>artificial bamboo forest> arid land.(3) The results of comprehend fuzzy-assessment is based on the format of membership function and index value, and the format depend on experts' experience and relationship between factors affecting soil fertility. The results demonstrate that: the IFI of paddy field, arid land, tea plantation, bamboo forest and nature woodland is 0.8407, 0.3731, 0.6431, 0.5301 and 0.7358, respectively, indicating that apparent difference display among soil with distinct utilization type(IFI of paddy field is the highest while IFI of arid land is the lowest). This can be described by the following order: paddy field> nature wood land> tea plantation>artificial bamboo forest> arid land.(4) We can draw a conclusion that the cerebellar model arithmetic computer successfully synthesized the index of specimen and evaluated them, which is objective. Although single item of indexes of one or two sample are significantly higher thatn others, it can make scientific evaluation results because it give consideration to all content of factors, and deplete interference, as well as analyze objectively. Therefore, the neural network model is more precise and objective than comprehend fuzzy-assessment model.