Fishing capacity evaluation of fishing vessel based on cloud model

In the evaluation problem of fishing vessel fishing capacity, the imperfect evaluation index system and the methods of evaluation indexes are mostly artificial qualitative evaluation methods, which lead to strong subjectivity and fuzziness as well as low accuracy of evaluation results. Therefore, this study introduces cloud model theory on the basis of improving the evaluation index system, converts the artificial qualitative evaluation results into the digital characteristics of clouds, realizes the mutual transformation of qualitative evaluation and quantitative evaluation, and improves the accuracy of evaluation results. Taking the trawler as an example, the cloud model method is used to evaluate the fishing capacity, and the result obtained is (77.1408, 1.6897, 0.0), the result obtained by the fuzzy comprehensive evaluation method is 76.664785, and the result obtained by the cloud center of gravity evaluation method is 0.7919. Compared with the other two methods, the cloud model method uses three numerical characteristics to describe the results, and combining the different numerical characteristics meanings, the evaluation results can be judged to be accurate, and the influence of ambiguity on the results is greatly reduced. Meanwhile, the evaluation results can be presented in the form of pictures, and the results are more intuitive; in addition, the cloud model of the evaluation results is compared with the standard cloud model for similarity, which improves the credibility and authenticity of the results.

www.nature.com/scientificreports/ vessels is obtained. The cloud drop distribution reflects the evaluation results of fishing capacity, and the weights of different indexes are considered when making cloud rule reasoning, which makes the evaluation results more reliable and objective.

Literature and research structure
Literature. From the previous summary, it can be seen that at present, the main research methods for fishing capacity of fishing vessels are DEA, stochastic frontier method and regression analysis, etc. The main research direction are fishery policy-making 5,6 ; surplus fishing capacity [7][8][9] ; the fishing efficiency of fishing vessels and fleets [11][12][13][14][15][16] ; the influence of technical and economic efficiency on fishing capacity 17,18 and related factors affecting fishing capacity [19][20][21][22][23][24][25] etc. At the same time, for the multi-condition evaluation problem, fuzzy comprehensive evaluation method and analytic hierarchy process are mainly used at present, which are difficult to eliminate or reduce the influence of fuzziness and subjectivity on the evaluation results. The above literature is summarized in Table 1.
Through the above analysis, most of the existing studies use DEA or regression methods to analyze the fishing efficiency and technical efficiency of fishing vessels or fleets, then help to formulate relevant policies. There is a lack of study on the strength evaluation of single-vessel fishing capacity, the relevant indicators used in the research are not comprehensive enough and the evaluation of single-vessel fishing capacity is a multi-condition evaluation problem. Therefore, this study combines previous studies with fishermen's experience to formulate a perfect evaluation index system of single-vessel fishing capacity. Using cloud model theory, combined with AHP and entropy weight method, the qualitative evaluation is transformed into quantitative expression, and the evaluation method of single ship fishing capacity is put forward. Comparing this method with cloud gravity center evaluation method and fuzzy comprehensive evaluation method, the feasibility of this method is verified, and the advantages of this method are shown.
Research structure. First of all, we should use the summary of past literature and the investigation of fishers and experts to improve the evaluation index system. Secondly, the cloud model theory combined with Table 1. literature summary.

Literature Main methods and brief content
Fishery policy and excess fishing capacity Grosskopf 5 Collier 6 Pham 7 Castilla-Espino 8 Quynh 9 Ji 10 DEA method is used to analyze fishery production capacity and fishermen, etc., so as to obtain the degree and reasons of excess fishing capacity, and formulate relevant management policies Fishing vessels and fleet different ship type fishing efficiency Gómez 11 Liang 12 Quijano 13 Van Hoof 14 Tsitsika 15 Tunca 16 DEA method is used to analyze fisheries, fishing vessels and fleets in different areas, and the factors affecting fishing efficiency and the influence of different ship types and nets on fishing efficiency are determined Technical efficiency and production efficiency Vazquez-RoweI 17 Li 18 Using DEA method to analyze the fisheries of different fleets and regions, it is found that the "captain effect" will really affect the fishing efficiency, and the lower technical efficiency will affect the development of fisheries Factors and indicators affecting fishing capacity Fang 19 Fang 20 Yajin 21 Ward 22 Chen 23 Xin 24 Damalas 25 By using DEA, standardized unit fishing effort catch(CPUE) and factor analysis, the fishing capacity of fishing vessels in different countries' sea areas and different types of operations was analyzed, and the relevant factors and indicators affecting fishing capacity were obtained Analytic hierarchy process, fuzzy comprehensive evaluation method Panchal 28 Kim 29 Li 30 Wu 31 Wang 32 Chen 33 Analytic hierarchy process (AHP) and fuzzy comprehensive evaluation method are classical evaluation methods, which can be used to evaluate the safety, disaster risk and reliability of parts Cloud model and its related improvement methods Yang 35 Wang 36 Xie 37 Li 38 Wang 39 Hou 40 Chen 44 Guo 45 Lü 46 Tan 47 Combining cloud theory with other methods, the S-shaped cloud model and Z-trapezoidal cloud model are derived. Combining with AHP and entropy weight method, the accuracy of cloud model evaluation is improved and applied in different fields Application of cloud model evaluation Zhao 41 Song 42 Du 43 Wang 49 Hou 50 Wu 51 Cloud model and related derivative methods are used to evaluate the development of power industry, power grid security and other aspects, and provide help for management. At the same time, this method is also applied to system efficiency evaluation and sustainability evaluation www.nature.com/scientificreports/ AHP and entropy weight method is used to form an evaluation method based on cloud model, and a single ship example is evaluated. Thirdly, fuzzy comprehensive evaluation method and cloud center of gravity evaluation method are used to evaluate the fishing capacity of the example. Finally, compare the three methods, and clarify the advantages and rationality. The specific method flow is shown in Fig. 1 below.

Evaluation index and weight calculation
Evaluation index system. The factors that have great influence on fishing capacity are total tonnage, total power, captain and fishing time [23][24][25] ,Based on the investigation results of experts and fishermen, four firstlevel evaluation indexes are determined as "fishing vessel specifications", "fishing technology", "net gear" and "resources and distribution of fishing objects". The first-level index contains 22 s-level indexes. The specific evaluation index system is shown in Fig. 2 below.
Index weight calculation. The traditional analytic hierarchy process (AHP) is easily interfered by subjective factors, which makes the evaluation result deviate 22 , so it can be regarded as subjective weight. Entropy method can determine an objective weight according to the fluctuation of data, to reduce the error caused by subjective factors 23 . The weight obtained by entropy method can be regarded as objective weight. Combining the two weights, the obtained weights are more reasonable, and both subjective and objective are taken into account.
Calculation subjective index weight. AHP can combine quantitative analysis with qualitative analysis, and establishes an orderly hierarchical structure, compares the weights of the upper elements of two layers, and comprehensively calculates the weights of the lower elements 51,52 .The specific steps are as follows 53 : Step 1: The objectives of the decision, the factors to be considered (decision criteria) and the decision options are stratified according to their interrelationship.
Step 2: Construct a judgment matrix and represent the elements in the judgment matrix A = (a ij ) n×n using a 9-bit scale.
Step 3: Single-level sorting and its consistency test. The maximum characteristic root of the matrix A = (a ij ) n×n is λ max , and the characteristic vector of λ max is marked as ω i after normalization, which is the subjective weight vector. The consistency test can be performed according to formula (1).
where CI is the consistency index, RI is the random consistency index and CR is the consistency ratio.When CR < 0.1, through consistency check, ω i can be used as a weight vector, otherwise, matrix A is reconstructed.
Calculation of objective index weight. The basic idea of entropy weight method is to determine the objective weight according to the variability of indexes.Entropy can not only reflect the degree of information confusion, but also measure the amount of information 53 .If the information entropy of an index is smaller, it indicates that the index is worth more variation, the more information it provides, the greater its role in comprehensive evaluation and the greater its weight.Specifically calculated by the following formulas.
(1) www.nature.com/scientificreports/ where x ij is the membership degree of the i-th object to the j-th index, x max and x min are the maximum and minimum values of the index respectively, P ij is the entropy information of the whole sample, H j is the entropy value, and ω j is the entropy weight.
Calculation of comprehensive index weight. Comprehensive weight ω can be determined by formula (5). Formula (5) combines subjective weight ω i with objective weight ω j , and α and β are the comprehensive proportions of subjective weight and objective weight in comprehensive weight 48 .

Evaluation methodology
Cloud model theory. Cloud theory related concepts. Cloud model is a model based on probability theory and fuzzy set theory, which transforms the qualitative concept into its quantitative representation through a specially constructed algorithm 54 .In the comment set X = {x}, the elements in it can map the comment set x to www.nature.com/scientificreports/ another ordered comment set X' according to a certain rule f. If there is only X' corresponding to x in x', then x' is the basic variable, and the distribution of membership μ in X' is called membership cloud 55 . The numerical characteristics of cloud are represented by Ex, En and He, Ex is the numerical value that best represents this qualitative concept in the comment set, En reflects the range that can be accepted by the concept in the comment set, He is the measure of entropy uncertainty, and reflects the randomness of samples of qualitative concept values. The three numerical features are represented in the cloud diagram as shown in Fig. 3.
Cloud generator. The forward path cloud generator is shown in Fig. 4a below. Normal cloud model can reflect the fuzziness and randomness of things or people's cognition in the objective world, and form a mapping between qualitative concepts and quantitative representations [54][55][56] .
The reverse cloud generator is shown in Fig. 4b below, the function of the reverse cloud generator is to find out the digital features Ex, En and He of the forward cloud generator according to the given cloud droplets, and to convert the quantitative representation into a qualitative concept.
Cloud rule generator. Cloud rule generator is a tool for uncertainty reasoning. Given the input, after activating the corresponding rule, it outputs the result. The cloud rule generator is composed of a preceding cloud generator and a succeeding cloud generator. The schematic diagram of the cloud rule generator is shown in Fig. 5.
The rules for uncertainty reasoning are as follows 56 : if A 1 , A 2 , …, A n then B.
Cloud model evaluation method. The evaluation method based on cloud model includes the following steps: Step 1: Building an evaluation index system and calculating relevant weights(The weight solution has been introduced in the third part); Step 2: Generate a standard comment cloud model; Step 3: Acquire expert scores and generate an expert score cloud model; Step 4: Use the cloud reasoning system for reasoning to obtain the evaluation result.  , Among them, "Excellent" and "Poor" are unilateral constraint comments, while "Good" and "Medium" are bilateral constraint comments, which can be described by one-dimensional normal forward cloud model.The digital characteristics of the forward cloud generator can be calculated by formulas (6)- (7).
In formula (6), the value range of constraint comments is [X -def , X max ] or [X min , X +def ], which is a unilateral constraint comment.In formula (7), the value range of constraint comments is [X min , X max ], which is bilateral constraint comments. According to experience, the value of k can be En/10, and the specific value should be combined with the actual situation.
Comprehensive cloud model for obtain expert comments. A rating range is used as input to the inverse cloud generator instead of the determined rating values, and the data features of the combined expert review cloud model are computed using each of the generated expert review cloud models 58 . Formula (8) can be used to calculate the comprehensive cloud numerical characteristics of n expert scores.
Building a cloud reasoning system. The cloud reasoning system consists of a cloud rule generator and a cloud rule base. After the system gets the input, a virtual cloud is generated by the cloud rule generator, and the virtual www.nature.com/scientificreports/ cloud is compared with the standard cloud model, and the comment corresponding to the cloud model with the highest similarity is the evaluation result. The specific workflow is shown in Fig. 6 below. The construction steps of the cloud reasoning system are as follows: Step 1: Build a cloud rule base and enter any value to activate the corresponding rule.
Step 2: A cloud computing rule generator is constructed, the weights are combined with an algorithm to decompose the multi-conditional cloud computing rule generator into several one-dimensional generators, and the output value of each one-dimensional cloud computing generator is calculated, and the weighted average method is used to obtain the final output 57 .
Step 3: Calculate the cloud graph similarity as a way to compare the similarity between the virtual cloud and the standard cloud model 59,60 .
The definition of similarity is as follows: Let two cloud images C 1 (Ex 1 , En 1 , He 1 ) and C 2 (Ex 2 , En 2 , He 2 ). If the membership degree of cloud drops (x i , μ i ) generated by C 1 inverse cloud generator in C 2 cloud is μ i ' , then the similarity between clouds C 1 and C 2 is 1 n n i=1 µ ′ i , which is recorded as δ 58 .
Cloud-gravity-center Assessing. Cloud-gravity-center Assessing uses the change of cloud gravity center to express the change of evaluation information. The change of cloud center of gravity reflects the change of information center, and the cloud center of gravity is expressed by formula (9).
where a is the position of cloud center of gravity and b is the height of cloud center of gravity. The Cloud-gravity-center Assessing is implemented as follows 35 : Step1: A cloud model is used to represent each indicator whose numerical characteristics are calculated by formulas (10)- (11).
where Ex and En are the expectation and entropy of the cloud model respectively, and x i is the data of sample. www.nature.com/scientificreports/ Step2: When the system changes, the integrated cloud model characterizing the system state also changes, and its center of gravity vector changes from Step3: Using the weights of the indicators, the weighted offset is calculated from formulas (12)(13)(14)(15).
where a 0 is the cloud center of gravity position vector in the ideal state, a is the cloud center of gravity position vector in the current state, b is the cloud center of gravity height, T i G is the normalized vector of T,T 0 is the cloud barycenter vector, ω j * is the weight of each index, and θ is the weighted deviation degree.
Fuzzy comprehensive evaluation method. The fuzzy comprehensive evaluation method converts qualitative evaluation into quantitative evaluation using the theory of affiliation, and the evaluation results are determined by the principle of maximum affiliation. The judgment matrix B and the final evaluation weight A are obtained according to the affiliation relationship, and the comment corresponding to the maximum value in vector B is the final evaluation result, which can be calculated by formulas (16)-(17) 32 . where the matrix R consists of the single-factor evaluation set single-factor evaluation set r i = {r i1 , r i2 , …, r in }, A = {a 1 , a 2 , …, a m } is the weight vector, a i is the weight of each factor, and B is the judgment matrix.

Case analysis
Take the single trawler * Yangjiang Fishing 0**8 as an example to evaluate the fishing capacity, and the comment set is V = {V 1 ,V 2 ,V 3 ,V 4 } = {Excellent, Good, Medium, Poor}. The specific information of the evaluation indicators is shown in Table 2 below. www.nature.com/scientificreports/ Tables 3, 4, 5, 6 and 7. The weights of the evaluation index system are also obtained by combining formulas (1)-(5) with expert scoring, as shown in Table 8 below.        Table 9, and the standard comment cloud model is shown in Fig. 7. Ten experts and fishermen scored the indicators, and the scoring results are shown in Tables 10, 11, 12 and 13. Since the selected calculation example is a single trawler, the indicators of gill net, gear, seine, fishing industry, net cover, and miscellaneous fishing gear do not need to be scored and the cloud model is (0,0,0), and the comprehensive evaluation cloud model of experts for each indicator is calculated by formula (8)    www.nature.com/scientificreports/ The obtained expert evaluation cloud model, combined with Table 9 and comment set, can activate relevant rules, taking fishing vessel specification evaluation as an example.Through the multi-condition single rule algorithm combined with weight, two cloud droplets are obtained as follows: x 1 = 84.75220327, x 2 = 69.52946857.

Determining index weight. The comparison matrix of each index is shown in the following
According to the inverse cloud generator algorithm, the numerical characteristics of the comprehensive evaluation cloud obtained from two cloud droplets are: (77.1408, 1.6897, 0.0).
According to the cloud image similarity calculation algorithm, the digital characteristics of comprehensive evaluation cloud are compared with those of standard cloud, and the obtained similarity is shown in Table 18 below.     www.nature.com/scientificreports/ According to Table 18, the virtual cloud should be between "Medium" and "Good", and the result is partial to "Medium". Through the forward cloud generator, the comprehensive evaluation cloud is drawn on the standard cloud-scale, and the evaluation cloud picture of fishing vessel specifications is shown in Fig. 8 below.
As can be seen from Fig. 8, the evaluation cloud picture is between the standard evaluation values of "Medium" and "Good", so the similarity with "Good" and "Medium" is higher than the other two evaluations. Among the two evaluations of "Medium" and "Good", the highest similarity is taken as the final evaluation of fishing vessel specifications, that is, the fishing vessel specifications are "Medium". When this method is used for evaluation, the evaluation cloud and the standard cloud map can be generated, and the similarity between the evaluation cloud and the cloud maps on both sides can be directly calculated. The larger similarity is the evaluation result.

Method comparison and result analysis.
Using the scores of experts and fishermen, the fuzzy comprehensive evaluation method and the "cloud-gravity-center" evaluation method were used to evaluate the fishing capacity of fishing boats. Taking the evaluation of fishing vessel specifications as an example, the affiliation vector of fuzzy comprehensive evaluation of fishing vessel specifications was obtained, and the evaluation results were obtained by inverse fuzzification. As shown in Table 19 below, the evaluation result is "medium" according to the principle of maximum membership. The weighted deviation θ = -0.2081 is obtained by "cloud-gravity-centered evaluation", and 0.7919 is input into the cloud generator, and the final activation result is "medium". The comparison results of the three methods are shown in Table 20 below, and the results of the three evaluation methods are the same, which verifies the feasibility of the method.
The comparison from Table 20 reveals that the results obtained from the cloud model contain three numerical features that can reflect the evaluation results more comprehensively, where "Ex" reflects the average level of the evaluation results, i.e., the current level of the evaluation object, and "En" reflects the dispersion of the cloud image, i.e., the reliability of the current evaluation results." The larger "En" is, the lower the reliability of the results, while "He" reflects the condition of the cloud drops 58 , and the evaluation results are described by three numerical features together, which improves the persuasiveness of the results. And the results obtained by the cloud computing method can be clearly pictorialized (e.g., Fig. 8), and each point in the image represents the quantitative concept transformed by the qualitative concept, and the accuracy of the conversion is reflected by the degree of certainty(vertical coordinate), and the pictorialized evaluation results make the evaluation results   www.nature.com/scientificreports/ more intuitive. At the same time, the cloud computing method also compares the evaluation results in terms of cloud map similarity, which further improves the accuracy of the results. Compared with the cloud model calculation method, the fuzzy comprehensive evaluation method can realize the transformation from qualitative to quantitative according to the principle of maximum affiliation, but its calculation process is more complicated, and the evaluation process completely relies on subjective scoring, and the results are often more subjective and fuzzy.
The cloud-gravity-centered evaluation method has a simpler calculation process and can better reduce the influence of subjective scoring on the results, but it only relies on a numerical feature to describe the evaluation results, which may make the results less accurate in practical application due to the one-sidedness and singularity of information.

Conclusion
Aiming at the problems in the evaluation process of fishing capacity of fishing vessels, such as incomplete evaluation indexes, mixed qualitative and quantitative descriptions, ambiguous indexes and substantial uncertainty, this paper puts forward a quantitative evaluation of fishing capacity of single vessel based on cloud model theory, and evaluates fishing capacity of single vessel by combining four first-class indexes such as fishing vessel specifications, fishing gear, fishing technology and resource distribution and corresponding second-class indexes. The research shows that: (1) The traditional qualitative inspection and scoring evaluation methods cannot accurately describe the evaluation object. In this study, the qualitative description of evaluation can be transformed into quantitative evaluation through scoring interval conversion and cloud model processing, and objective, accurate and unified quantitative evaluation results can be obtained. (2) The three numerical characteristics of the comprehensive evaluation cloud are obtained based on the cloud model theory, and the three numerical characteristics have different meanings, so it is able to consider the fishing vessel fishing capacity evaluation from both the quantitative results and the reliability of the results. For the decision evaluation problem with complex multi-attribute factors, the objectivity and accuracy of the evaluation results can be further improved. (3) Using the same-scale cloud images to directly reflect the evaluation results, the similarity comparison of cloud model provides objective basis for the cloud image results, and the cloud model theory is very suitable for solving the problem of fishing vessel performance evaluation covering multi-attribute uncertain factors. This study promotes the application of uncertainty information theory in the field of fishing vessel performance analysis, mining and evaluation decision engineering.
The evaluation of fishing capacity of fishing vessels depends on a complete and objective evaluation index system and an evaluation method that can minimize subjective influence. Perfecting the index system to describe fishing capacity of fishing vessels more accurately, and proposing a completely quantitative evaluation system that does not depend on subjective scores will be the next research work.

Data availability
The data come from the data on offshore fishing operations in China obtained from the China Fishery Statistical Yearbook published from 2004 to 2020.