Bayesian Network in the Consumption of Forest Products in the Quality and Safety

Abstract. The consumption of forest product quality and safety issues more and more attention of the society, only analysis of the impact of the consumption of forest product quality and safety factors, combined with the traceability process, establish and improve the traceability model, in order to realize the quality and safety monitoring of the consumption of forest products. To do this, the establishment of a bayesian network based edible forest product quality safety traceability of the initial model, the model can harmonize "prior" prevention and "post" track, in order to protect the quality and safety of the consumption of forest products.

Key words: Consumption of Forest Products; quality and safety; Retrospective

Introduction

Edible forest product quality and safety related to the image of the government and social stability is related to forester’s income, regardless of any time we cannot ignore the major issues [1]. Forest products, food production, circulation and consumption process more complicated, and our government used in the food quality is the multispectral division of regulatory mode [2], It is difficult for food sources of forest products for the accurate and reliable retrospective. This is necessary to establish a sound advantage retrospective model, to ensure that food from the forest to the table all aspects of the information can be traced back.

Edible forest product quality and safety traceability system are a network structure, in order to make the system more good you have to pick the right model. Loose [3] to describe the food traceability models Gozinto graphs; CA van Dorp [4] also has been improved, and proposes a model to achieve. There are many problems in the food traceability system: dispersion problem, by Gozintographs described to solve this problem, but this problem is NP use Gozinto graphs still need to improve and refine. The Bayesian network combines the knowledge of graph theory and probability theory, provides a compact graphical method for the expression of complex probabilistic uncertainty between random variables, there is a great advantage in solving NP problems. Therefore, planting the source grabbed from the consumption of forest products, identify the key factor in the consumption of forest products in the production process affect the quality and safety of these key factors as critical control points, combined with production processes, the use of Bayesian network theory to establish the traceability model, so that the "prior" prevention and "after" tracking.

Analysis of the Consumption of Forest Products Quality and Safety of Influencing Factors

Consumption of forest products from forest to table "process, go through the production of raw materials, transport storage, processing, production, marketing and distribution, and consumer products supply chain process, the factors that affect the quality and safety will be the impact of every aspect of [5 ]. Combined with the author's research progress and work practices, based on extensive research and seek expert advice, forest products affect food quality and safety factors into the environment for the growth, production, processing, circulation sales three aspects. Among them, production and processing stages of food production safety of the forest products play a decisive role. Stages of production and processing for the acceptance of raw materials, raw material storage, processing, product packaging and other links, each link will bring one or more of the final product quality and safety factors [6], and business management methods and staff awareness have an impact on product safety. Due to frequent natural disasters in recent years, causing a variety of diseases and insect pests of forests, increased use of pesticides, pesticides and improper use of unqualified, the most important factor is the impact of the consumption of forest products safety [7].

In summary, the combination of food every link in the supply chain for forest products harm the possibility of the occurrence and its cause, you can come to the impact of the initial consumption of forest product quality and safety of the main factors, namely, air quality, soil quality, irrigation water quality, pest management, fertilizer management, process management, transportation management, sales management, and so on.

Forest product quality and safety will be affected by these factors, inextricably close relationship between them, and this relationship is dynamic, as one of the factors of change and change, and ultimately affects the consumption of forest product quality and safety. So, as soon as possible to shape the consumption of forest product quality and safety traceability model to better show the relationship between these factors and the dynamic changes, so that in a good grasp of these dynamic changes from a factor of change know the trend of consumption of forest product quality and safety of the ultimate impact of the results; You can also find the various aspects of the problems and their causes, remedies and correction in the shortest possible time, be able to minimize the loss and the impact of other factors.

Analysis of the Edible Forest Products Retrospective Process

In the food supply chain, when food is passed from one node to the next level node, can be used for product traceability information will be stored in the event of a food safety retrospective. Growers, processors and marketers of the food supply chain are responsible for the collection of data. They collect data, such as product identification data and other business process information is sent to the database of the Food Traceability System [8]. At the same time, these data can be used to establish an internal information system. In this way, the continuous flow of products and trace the flow of information to constitute a complete supply chain. All food-related information is transparent and traceable [9].

Since most of the food traceability systems have the same basic features, we can proceed from a typical food traceability process through a retrospective model simplification.

Consumption of Forest Products Quality and Safety Traceability Model

Bayesian network (Bayesian Networks), also known as belief networks (Belif Networks) or causal network (Causal Networks), indicates that the variable dependency between the probability of a directed acyclic graph, where each node represents a random variable, each probability dependence edges represent variables, each node corresponds to a conditional probability distribution table (CPT), indicating the probability of the number of dependent relationship between the variable with the parent node.

Definition 1 Bayesian structure: a set of variables The Bayesian network is composed of two parts: a network structure S U, independent variable conditions assertion. S is a directed acyclic graph; the nodes in Figure 1, and you in the variables correspond to the diagram to the edge expressed condition dependence between variables. Second, is the conditional probability distribution associated with each of a variety P [10]. S and P define a U joint probability distribution.

According to the definition of the structure of the Bayesian network, the steps for building a Bayesian network:

1) Identified as model variables and their interpretation.

2) The establishment of a conditional independence assertions directed acyclic graph. Usually two steps: ? determine the order of the nodes; ? decide each node's parent. ? It is the most critical step. The Bayesian network structure contains all the variables of the joint probability distribution, and performance in the form of conditional probability that

Indicates the parent node set of X i.

Otherwise, in the case of a given parent node, X and the other nodes are independent, i.e.

For each variable X, there is a subset of make and is Conditional independence, that is, for any X,

From the above equation, the variable collection and corresponding, that is, the process of determining the parent node by looking for the conditions between the variables independent relationship is completed to determine the conditional independence relations will determine each node's parent.

Assigning the conditional probability distribution for each node . Correct network structure cannot be obtained because the order of the variables may be wrong. In order to get the correct result, the need to consider up to Species in a different order, laborious. Taking into account the: ? variable between causality easily found; ? causality and conditional independence assertions corresponding to. Therefore, we use another method to build Bayesian networks: Bayesian network structure from the reason variable, which can be drawn between the direct results of using a connection to the arc.

Expert evaluation and historical data fusion Bayesian network structure-modeling method. Structure is determined; the next task is to determine the conditional probability distribution before the data library statistics obtained can also let the experts determined empirically. Tedious complex problem, the more variables involved, the harder it is to determine the relationship between the variables, the conditional probability distribution is difficult to determine. Therefore, many researchers have studied a method of learning Bayesian networks from data, and to shape the model instead of a human expert. From the data, learning Bayesian network can save the labor of experts, objective results, the drawback is the slow convergence of the algorithm, and the learning process requires a large amount of sample data support, if the data is less, the result is not necessarily accurate. Combine data and expert knowledge, will both learn from each other is a relatively correct modeling idea. There to propose a modeling method, shown in Figure 1.

Fig.1. Expert evaluation and historical data integration Bayesian networks structure modeling

The specific method is as follows: Let Bayesian network a total of n variables, each pair of variables by evaluation experts to determine the causal relationship, so that a total of Right. To avoid the singularity of such an expert, you can adopt the weights analysis a number of experts views. Can directly determine the causal relationship between all the variables and not some time judgment is not entirely correct. Obviously, a single virtue of expert advice and cannot get the final result. I use the evaluation method combined with sample data, already rely on the opinions of experts to determine a causal relationship between most of the variables, so that the basic structure of the network can be to be identified, then the use of the data before the data library the completion of the remaining structure.

Edible forest products quality and safety traceability model based on Bayesian network. According to previously know, if it is already known to affect the consumption of forest products quality and safety of the main factors can be initially identified system variables. Order to better refine this model, the factors are management areas, such as the processing, marketing, transportation management, unified production management Ultimately determine the system variables in the model: air quality, soil quality, irrigation water quality, pest management, fertilizer management, production management, and consumption of forest products quality. According to experts suggest a causal relationship table, as shown in Table 1. The table to the right of the arrows indicates the line attribute column properties of the parent node to the left of the arrow indicate the column attribute line properties of the parent node. The bi-directional arrows indicate the relationship cannot be determined, a blank indicates no relationship between the two. Whether the level of production management can directly affect the quality and safety of edible forest products, the experts are not sure of their relationship, and can be combined with statistics inferred. A forest products company in Jilin Province, for example, the standard basis for processing management, transportation management, production safety management, production management level is divided into excellent, qualified and non-qualified three levels. By the local safety inspection and testing center can show that the detection data of the past three years, the consumption of forest products quality and safety directly affected by the impact of the production management of the level of the initial establishment of a Bayesian network model shown in Figure 2.

Fig.2. Consumption of forest products quality and safety traceability Bayesian network model of initial

Based on historical statistical data, may determine the conditional probability distribution of each node in the network, to consumption of forest products quality junction and the production management level cite cases, its conditional probability table shown in Table 2.

Table 1. Consumption of forest products quality junction conditional probability tables (part)

Production management level Consumption of forest products quality

Excellent Qualified Fail

Qualified 9.1 8 8.7 6 2.6

Unqualified 0.9 1.3 7.4

Implementation of appropriate measures if it determines the level of production management to be ineligible, the evidence of this condition as an input Bayesian network model, the consumption of forest product quality node for failing posterior probability will be up to 7.4%. You can increase the number of test products to defense failed to enter the market. In the model if there is evidence of some nodes, such as consumption of forest products, the quality of existing problems, but also know the various pollution detection does not exist, you can draw the posterior probability of the other nodes in the condition, so that you can be from the most began to link to test.

Conclusion

A large number of factors can affect the quality of edible forest products safe, indivisible close links between these factors; use of Bayesian networks describes the consumption of forest products quality and safety traceability model to more accurately reflect the causal links between these factors and changes relationship. Bayesian networks are more difficult to establish, especially in the face of a more complicated problem, involving many variables, the relationship between variables is difficult to decide to combine data and expert knowledge, so that they are able to learn from each other is a more feasible modeling idea.

References

[1] Pan Bingliang, the leaves long, Huarui Fu measures to increase the consumption of forest products safety work to explore [J] the green Times, 2012 (8): 86-87

[2] Chen Hsing Gang, Zhang Hailin the forest food safety problem [J]. China Forest specialties 2008 (4): 88-89

[3] R. Ramsower, JW indsor, JI Induction of a reference data model for tracking and tracing [C] / / Proceedings of the 8th Americas Conference on Information Systems (AMCIS 2002), 2002

[4] DeGross Association for Information Systems [M] USA, 2002: 945-957

[5] Xiao Jing. Based on the food security of the supply chain study [D]. Changchun: Jilin University, 2009

[6] Yang Yantao processed agricultural products quality and safety warning and empirical research [D]. Beijing: Chinese Academy of Agricultural Sciences, 2009

[7] Luo Liang, YUE Yong Tang Feng, Liu Chuanqin quality and safety problems of China's consumption of forest products briefly [J] Anhui Agricultural Science Bulletin 2008,14 (14): 25-26

[8] Gao Rong, Research-based the pork traceability and the price early warning model of Things [D]. Chengdu: University of Electronic Science and Technology of China 2011

[9] Cheng Yu China's food quality and safety traceability system construction [J] TRADE INDUSTRY, 2012, 20: 33-34

[10] Hu Xiaoxuan. Bayesian network modeling technology and its application in decision-making [D] Hefei: Hefei University of Technology, 2006