A study on bilateral matching of team-science talents in new R&D institutions based on grey correlation-cloud model

2.3.1 Characterization of T-S assessment values in terms of interval grey numbers

The T-S assessment value is the specific data given by the new R&D institution team T to the scientific and technological talent S based on the T-S assessment index. The team of the new R&D organization gives a percentage score when assessing the scientific and technical talent. Since different team representatives are involved in the assessment, there are assessment differences in these representatives’ own learning, preferences, reference points, information asymmetry, etc., so the assessment results are not exactly the same. However, since they belong to a new type of R&D organization, the head of the organization often has some introduction to the assessment, so that the difference between the assessment results will not be too large, and the assessment data is a range of grey numbers, which do not know the specific value of the assessment, but clearly know the range of the assessment value. 1982, our scholar Professor Deng Julong proposed the “grey system theory”, which is to solve this kind of uncertainty of small sample data set problem. Grey number is the basic “cell” of grey system theory, which is an uncertain number of values in a certain interval, only knowing the range of values of information but not the exact value of information, usually denoted by the symbol ⊗. A grey number that has both a lower and an upper bound is called an interval grey number, denoted as ⊗ ∈ [ a _ , a ¯ ] , with a _ denotes the lower bound and a ¯ denotes the upper bound. Using the interval grey number to characterize the T-S assessment values, we obtain the decision matrix A = [ a ij p ] , i = 1 , 2 , ⋯ m , j = 1 , 2 , ⋯ k , p = 1 , 2 , ⋯ n , i denotes that there are m scientific and technical talents, j denotes that there are k T-S assessment indicators, and p denotes that there are n teams. A = [ a _ 11 p , a ¯ 11 p a _ 12 p , a ¯ 12 p ⋯ a _ 1 k p , a ¯ 1 k p a _ 21 p , a ¯ 21 p a _ 22 p , a ¯ 22 p ⋯ a _ 2 k p , a ¯ 2 k p ⋯ ⋯ ⋯ ⋯ a _ m 1 p , a ¯ m 1 p a _ m 2 p , a ¯ m 2 p ⋯ a _ mk p , a ¯ mk p ]

2.3.2 Normalization of interval grey number assessment values based on the maximum value method

The maximum value transformation method is intuitive, convenient, and fast to achieve the normalization requirements while preserving the characteristics of the original data, and is one of the most commonly used transformation methods for grey numbers in multi-attribute intervals. In this paper, the maximum value transformation method is used to normalize the T-S assessment values. The maximum value transformation method uses the initial evaluation values of a _ ij p and a ¯ ij p as the numerator, the maximum assessed value under each indicator a ¯ j as the denominator, and dividing the two to obtain the normalized T-S assessment value B = [ b ij p ] = [ b _ ij p , b ¯ ij p ] . b _ ij p = a _ ij p max ( a ¯ j ) , b ¯ ij p = a ¯ ij p max ( a ¯ j ) i = 1 , 2 , ⋯ , m ; j = 1 , 2 , ⋯ , k(1)

Definition 1. Let the interval grey number ⊗ ∈ [ b _ , b ¯ ] , the background or domain of its generation as rmOmega, define ⊗ ˆ = E ( ⊗ ) be the kernel of this interval grey number. In the absence of information on the distribution of the values taken, if the interval grey number is a continuous grey number, it is said ⊗ ˆ = 1 2 ( b _ + b ¯ )(2) for the kernel of the grey number in this interval [22]. l ( ⊗ ) = b ¯ - b _(3) is the step size of the grey count for that interval.

Definition 2 [22]. Let two grey numbers ⊗ 1 ∈ [ b _ 1 , b ¯ 1 ] , ⊗ 2 ∈ [ b _ 2 , b ¯ 2 ] , define σ ( ⊗ 1 , ⊗ 2 ) = | ⊗ ˆ 1 - ⊗ ˆ 2 | + 1 2 | l ( ⊗ 1 ) - l ( ⊗ 2 ) |(4)

call σ (⊗ ₁, ⊗ ₂) as the distance between ⊗₁ and ⊗₂. Where | ⊗ ˆ 1 - ⊗ ˆ 2 | denotes the distance between two interval grey number kernels, and 1 2 | l ( ⊗ 1 ) - l ( ⊗ 2 ) | denotes the distance between the lengths of the two interval grey number semi-intervals.

In 1982, Professor Deng Julong published a paper on “Control Problems of Grey Systems” and “Grey Control Systems”, which marked the beginning of the new cross-cutting discipline of grey system theory [34]. The research object of grey system theory is a “small sample” and “information-poor” uncertain system with “partly known information and partly unknown information”. It understands the real world through the generation and development of “partially” known information, and realizes the correct grasp and description of the system operation behavior and evolution law. In literature [35], a fast typical target recognition algorithm f small samples is proposed, which uses a laser line scan model to complete the two-dimensional scanning of the target and obtains the distance value of the target to be recognized, forming the sample sequence matrix of the recognition algorithm, thus providing a reference for future applications of low-cost laser imaging fuse target recognition. The literature [36] proposed a top-down prediction method based on grey system theory to reveal the future trend of annual heat supply to assist urban planning and energy structure reform. The literature [37] applies the grey system theory MGM (1,4) to predict the development of road traffic accidents in Germany until 2025 based on the market diffusion of electronic stability plans (ESP). The literature [38] proposes a new supplier selection tool that combines grey systems theory with uncertainty theory without any probability distribution nor fuzzy affiliation function to not only select the most suitable supplier but also to assign the optimal purchase quantity under stochastic and identifiable uncertainty. The literature [39] proposed the grey point assignment full consistency (grey-PA-FUCOM) weighting method and the extended grey state method based on grey system theory to solve the uncertainty in the process of evaluating suppliers. From the above literature, it can be seen that: the main contents of grey system theory include grey correlation analysis, grey series generation, grey prediction model, etc. There are no special requirements and restrictions on the observation data and its distribution, and it has very wide application fields for system analysis, evaluation, prediction, decision making, control, optimization, etc.

The basic idea of grey correlation analysis is to determine whether the series curves are closely related according to their similarity in geometry. The closer the curves are, the greater the degree of correlation between the corresponding series, and vice versa, the smaller the crelation. The grey correlation analysis method is equally applicable to the amount of sample size and the presence or absence of aattern in the sample, and it is very convenient with a small calculation, and there is no discrepancy between the quantitative results and the qualitative analysis results. The literature [40] conducted a comprehensive evaluation owaste heat recovery technologies for offshore alications based on hierarchical-grey correlation analysis, analyzed the correlation between each cost and studied the impact of individual costs on the total cost, which helps to determine the most suitable heat recovery technology for marine engines. The literature [41] proposed 16 potential factors influencing carbon emissions and used grey correlation analysis to identify factors with strong correlation with carbon emissions, providing a theoretical basis for reducing carbon emissions in China. The paper [42] analyzed the relationship between the correlation coefficient and the physical properties of recycled coarse aggregate (RCA) based on the grey correlation analysis (GCA) method to achieve the prediction of the compressive strength of recycled aggregate concrete. The literature [43] proposed an adaptive kernel principal component analysis (AKPCA-GRA) method combined with grey correlation analysis (GRA) to dynamically monitor the occurrence of failures. In the literature [44], the safety of nonlinear hydro generator units (HGU) operating under different loads was analyzed using a combination of entropy weight method and grey correlation analysis. From the above literature, it can be seen that grey correlation analysis can be used for assessment, influence factor analysis, prediction, monitoring of faults and safety, etc. It can also be combined with different methods, such as hierarchical analysis, principal component analysis, entropy weight method, etc., showing better problem solving capabilities.

In this paper, the maximum value is used as the reference point. Due to the existence of the reference point, the team of the new R&D organization and the scientific and technical personnel care more about the difference between the result and the ideal solution rather than the result itself. The closer the assessed value under a certain indicator is to the maximum value, the higher the satisfaction of the decision maker under that indicator, and the farther the distance from the maximum value, the lower the satisfaction under that indicator. In this paper, we calculate the grey correlation coefficient based on grey correlation analysis and get the T-S correlation coefficient matrix C = [ c ij p ] . The calculation method is as follows.

Definition 3 [34]. Let the system sample sequence matrix be B = [b_ij], the maximum reference sequence B + = ( b i 1 + , b i 2 + , ⋯ , b in + ) , and for ξ ∈ (0, 1), such that α ij + = min i min j σ ( b ij , b ij + ) + ξ · max i max j σ ( b ij , b ij + ) σ ( b ij , b ij + ) + ξ · max i max j σ ( b ij , b ij + )(5)

σ ( b ij , b ij + ) is the distance between b_ij and b ij + . The distance between b_ij and b ij + is calculated using the interval grey number distance formula. The grey-off linkage number has four characteristics: normality, wholeness, even-pair symmetry and proximity. Normality indicates that the grey-off linkage number is between 0 and 1, i.e. 0 < r_ij ⩽ 1, indicate that no two behavioral sequences in the system can be strictly uncorrelated. Holisticity reflects the influence of the environment on the comparison of grey correlations, which varies with the environment. Even-pair symmetry shows that symmetry is satisfied when there are only two sequences in the set of grey correlation factors. Proximity is a constraint on the quantification of correlations.

Let the T-S subjective indicator weights be ω_1j, the T-S objective indicator weight is ω_2j, and the T-S composite indicator weight is ω_3j. The objective indicator weights are calculated based on the grey correlation coefficient ω 2 j = ∑ i = 1 m ∑ p = 1 n c ij p ∑ j = 1 k ∑ i = 1 m ∑ p = 1 n c ij p(6)

Introducing subjective-objective weighting parameters θ, the weights of the subjective indicator weights are θ, the weights of the objective indicator weights are 1 - θ, then the combined indicator weights, ω 3 j = θ * ω 1 j + ( 1 - θ ) ω 2 j(7)

The matrix of T-S grey correlation coefficients C = [c_ij] according to the combined indicator weights ω_3j. The clustering is performed. In order to keep the normality of the assembled data and the original correlation coefficient information from being lost, and to consider the integrated indicator wghts, this paper takes 1/max(ω_3j) and performs the reconciliation tobtain the new T-S evaluation correlation coefficient matrix D = [ d ij p ] . d ij p = c ij p × ω 3 j × ( 1 / max ( ω 3 j ) )(8)

The cloud model was proposed by Professor Li Deyi in 1995 to deal with the uncertainty transformation of qualitative concepts and quantitative descriptions [27], which breaks the traditional fuzzy assessment method and abandons the practice of using the exact affiliation function to calculate the median value of the thesis domain, followed by the expectation “Ex”, entropy “En” and super entropy “He” to reflect the uncertainty of the concept. The cloud model can be converted from qualitative to quantitative by forward cloud generator, which generates cloud drops based on the numerical characteristics of clouds. The cloud model, through the inverse generator, then converts the exact data into a qualitative concept represented by numerical features.

Definition: let U be a quantitative domain represented by exact values and C be a qualitative concept on U. If the quantitative value x is a random realization of the qualitative concept C, the degree of certainty of x with respect to C. μ _x  ∈ [0, 1] is a random number with a tendency to be stable. μ : U → [ 0 , 1 ] ∀ x ∈ U x → [ 0 , 1 ]

Then x in the domain of the theory U. The distribution over a cloud is called a cloud and x is a cloud drop.

“Ex” is the mathematical expectation in the theoretical domain. “En” is the entropy, which represents the uncertainty of the qualitative concept, and the entropy E_n the greater the uncertainty of the expectation, the higher the uertainty of the expectation. “He” is the entropy oentropy, which indicates the uncertainty of entropy, usually reflected in the thickness of the cloud, where the super entropy “He” the greater the entropy, the thicker the cloud. The tee numerical characteristics of the cloud model are shown in Fig. 2.

OPEN IN VIEWER

Fig. 2

Nmerical characteristics of the cloud model.

The cloud model expectations of the tth team for the sth talent are Ex ts TS = 1 k ∑ j = 1 k d ij p(9)

The entropy of the cloud model expectation of the tth team for the sth talent is En ts TS = π 2 1 k ∑ j = 1 k | d ij p - Ex ts TS |(10)

The hyperentropy of the cloud model expectations of the tth team for the sth talent is

He ts TS = S 2 - En ts TS 2 = [ 1 k - 1 ∑ i = 1 n ( d ij p - Ex ts TS ) 2 ] - En ts TS 2(11)

In this paper, we invert the cloud generator by cloud model to transform the decision matrix D = [ d ij p ] transformed into a TS cloud data matrix,

TS = [ Ex ts TS , En ts TS , He ts TS ] t = 1 , 2 , ⋯ , n ; s = 1 , 2 , ⋯ , m .(12)

2.4.1 Characterizing S-T assessment values with a cloud model

The S-T assessment value is the specific data given by the scientific and technical talent S to the new R&D organization team T based on the S-T assessment index. The language assessment value given by the new R&D organization team in the assessment of the new R&D organization by the S-T talent. Since the language assessment itself also has some uncertainty, the standard cloud of the cloud model is applied to convert the language assessment values into cloud data.

In this paper, we divide the theoretical domain [0,1] into five evaluation levels based on the golden partition method to establish the cloud model scale for S-T language evaluation. Let 0.618 be the multiplier between the cloud model parameters of adjacent levels, and the closer to the central region of the thesis domain [0,1], the smaller the entropy and hyper entropy of the assessment levels. Taking the focus 0.500 of the theoretical domain [0, 1] as the intermediate assessment level, the cloud model parameters of E_x = 0.500, the H_e = 0.005, with specific values as in Table 3.

Based on the transformation scale, the S-T expectation decision matrix is obtained E = [ e ij p ] , i = 1 , 2 , ⋯ n , j = 1 , 2 , ⋯ r , j = 1 , 2 , ⋯ m , e_ij is the expected value of the transformed cloud data, i indicates that there are m talents, j indicates that there are r S-T assessment indicators, p indicates that there are n scientific and technical talents. E = [ e 11 p e 12 p ⋯ e 1 r p e 21 p e 22 p ⋯ e 2 r p ⋯ ⋯ ⋯ ⋯ e n 1 p e n 2 p ⋯ e nr p ]

Sce the expected values of cloud data meet the normalization requirements, the initial assessment values do not need to be normalized again. In this paper, the maximum value of the expected value of the initial assessment value is used as the reference point, and the grey correlation coefficient is calculated based on grey correlation analysis to obtain the T-S correlation coefficient matrix F = [ f ij p ] . The calculation method is as follows. β ij + = min i min j σ ( e ij p , e ij p + ) + ξ · max i max j σ ( e ij p , e ij p + ) σ ( e ij p , e ij p + ) + ξ · max i max j σ ( e ij p , e ij p + )(13)

σ ( e ij p , e ij p + ) is the expected value e ij p and the maximum value of the expected value e ij p + distance between them. The closer the assessed value under an indicator is to the maximum value, the higher the satisfaction of scientific and technical personnel with that indicator. The farther the distance from the maximum value, the lower the satisfaction of scientific and technical personnel with that indicator.

Let the S-T subjective indicator weights be ω_4j, the S-T objective indicator weight be ω_5j, and the S-T composite indicator weight is ω_6j, this paper calculates the objective indicator weights based on the grey correlation coefficient ω 2 j = ∑ i = 1 m ∑ p = 1 n f ij p ∑ j = 1 r ∑ i = 1 m ∑ p = 1 n f ij p(14)

Introducing subjective-objective weighting parameters δ, the weights of the subjective indicator weights are δ, the weights of the objective indicator weights are 1 - δ, then the combined indicator weights ω 6 j = δ * ω 4 j + ( 1 - δ ) ω 5 j(15)

The matrix of T-S grey correlation coefficients F = [ f ij p ] are according to the combined indicator weights ω_6j. The clustering is performed. In order to keep the normality of the assembled data and the original correlation coefficient information from being lost, and to consider the integrated indicator weights, this paper takes 1/max(ω_6j) and performs the reconciliation to obtain the new T-S evaluation correlation coefficient matrix G = [ g ij p ] . g ij p = f ij p × ω 6 j × ( 1 / max ( ω 6 j ) )(16)

The cloud model expectations of the sth talent for the tth team are Ex ts ST = 1 r ∑ j = 1 r g ij p(17)

The entropy of the cloud model expectation of the sth talent for the tth team is En ts ST = π 2 1 r ∑ j = 1 r | g ij p - Ex ts ST |(18)

The super entropy of the cloud model expectation of the sth talent for the tth team is

He ts ST = S 2 - En ts ST 2 = [ 1 r - 1 ∑ i = 1 n ( g ij p - Ex ts ST ) 2 ] - En ts ST 2(19)

In this paper, we invert the cloud generator by cloud model to transform the decision matrix D = [ d ij p ] transformed into ST cloud data matrix, ST = [ Ex ts ST , En ts ST , He ts ST ] t = 1 , 2 , ⋯ , n ; s = 1 , 2 , ⋯ , m .(20)

2.5.1 Bilateral matching theory

Professor Gale of Brown University and Professor Sharply, a renowned economist, published the paper “College Admissions and Stable Marriages” in 1962, which studied match existence, the Gale-Sharply algorithm, Pareto optimality, and stable matching, and pioneered the theory of bilateral matching decisions. Professor Roth of the University of Pittsburgh introduced the concept of bilateral matching explicitly in 1985.

Let μ : P ∪ Q → Q ∪ P be a one-to-one mapping, if ∀P_i ∈ P, ∀ Q_j ∈ Q, satisfies μ (P) _i = Q_j ∈ Q and μ (Q_j) = P_i ∈ P, then we call μ is a two-way matching. μ (P_i) = Q_j denotes P_i with Q_j in μ is in the match, and μ (Q) _j = Q_j indicates at Q_j in μ does not match.

For real-life bilateral matching decision problems, forming a reasonable and effective bilateral matching solution is a common need for all subjects, and constructing a bilateral matching model and proposing a solution algorithm is the most critical step. The traditional Gale-Sharply algorithm is based on male-optimal (Man-optimal) or woman-optimal (Woman-optimal), and the Hospital-Resident aorithm is based on hospital-optimal (Hospital-optimal) or intern-optimal (Resident-optimal) can only obtain a bilateral matching solution that is satisfactory for some subjects and does not guarantee the fairness of bilateral subjects.

With the development of market economy, the team of new R&D institutions and scientific and technological talents are on an equal footing, and the matched subjects are influenced by individual characteristics, the internal and external environment they are in, desired goals, geographical differences and their own preferences, giving assessments of the other subject that are not necessarily preference-order information, but may be linguistic information, uncertain information, etc.

2.5.2 Combination of cloud models

The different forms of uncertain initial assessment values have been assembled above to obtain two sets of cloud model representations of the assessment values, each with three numerical features. The TS matrix is the assessment value of the new R&D organization team T on the technological talent S. The TS matrix is the assessment value of the technological talent S on the new R&D organization team T. The TS matrix is the assessment value of the technological talent S on the new R&D organization team T. The TS matrix is the assessment value of the new R&D organization team T.

When the cloud data is combined and integrated, all 3 digital features need to be combined and operated as follows.

Definition 4. let H_i = (Ex_i, En_i, He_i) (i = 1, 2, ⋯ n) be a set of clouds corresponding to a weight vector ρ = (ρ₁, ρ_n, ⋯ ρ_n) , ρ_i ∈ [0, 1] (i = 1, 2, ⋯ n), ∑ i = 1 n ρ i = 1 , then the set of n clouds is still a cloud whose cloud-weighted arithmetic average operator CWAA is

CWAA ( H 1 , H 2 , ⋯ H n ) = ∑ i = 1 n ρ i H i = ( ∑ i = 1 n ρ i Ex i , ∑ i = 1 n ρ i En i 2 , ∑ i = 1 n ρ i He i 2 )(21)

Expectation is the most important feature among the 3 numerical features of cloud data, and its magnitude determines the final sign of whether the new R&D organization or the scientific and technological talent is satisfied or not. Since there exists a certain degree of importance for the new R&D organization team and the scientific and technological talents, we introduce the importance parameter ρ to indicate the matching dominance of the new R&D organization team. For example, if the new R&D organization team dominates more in the matching than the scientific and technological talents, then ρ > 0.5. If the tech talent dominates more in the match, then ρ < 0.5. If team and talent are equally important, then ρ = 0.5. We follow the dominance parameter ρ by combining the expected value Ex ts TS and Ex ts ST . Perform the set to obtain the combined expectation Ex_ts. Ex ts = ρ × Ex ts TS + ( 1 − ρ ) × Ex ts ST t = 1 , 2 , ⋯ n , s = 1 , 2 , ⋯ m(22)

Based on the cloud integration formula and the team-talent dominance combination formula, we obtain the combined cloud data matrix for both parties’ satisfaction W = [Ex_ts, En_ts, He_ts].

The combination expectation value Ex_ts is the larger the better. We consider the tth team and the sth talent to be matched if the combination expectation value reaches the maximum. The entropy En of the assessment value represents the uncertainty of the expectation value, and we want the assessment value to be as close to the expectation value as possible, so the smaller the uncertainty the better, then the smaller the entropy En of the combination.

In this paper, we develop a bilateral matching model M-1 based on bilateral matching theory and cloud data matrix W. MAX Z = ∑ t = 1 n ∑ s = 1 m π ts ( 1 - En ts ) Ex ts s . t . { π ts ∈ [ 0 or 1 ] ∑ t = 1 n π ts = 1 ∑ s = 1 m π ts ⩾ 1 ( M - 1 )(23)

The objective of the M-1 model is to transform the bi-objective with maximum expectation and minimum entropy into a single objective. The constraints are: the weight vector π_ts can only be 0 or 1; each tech talent must be matched to a new R&D organization team, and only 1 team can be matched; each team must be matched to a tech talent, and multiple tech talents can be matched.

Theorem 1. The model M - 1 must have an optimal solution.

According to the optimality existence theorem, any single-objective programming with bounded feasible domain must be optimal over its feasible domain. The model M - 1 is a single-objective programming problem and the feasible domain of the model exists and is bounded, so the M-1 model must have an optimal solution. M - 1 can be solved by LINGGO software, and the solution of π_ts = 1 means that the tth team of new R&D organization and the sth technological talent are successfully matched.

3.2.1 Collect assessment data and characterize the initial assessment values in terms of interval grey numbers

In this paper, we collected the assessment interval values of six scientific and technological talents based on 10 assessment indicators from teams T1-T4 in the new R&D organization, as shown in Tables 4–9. There are 12 science and technology teams in the new R&D organization T. Four of the teams have similar specialties, and each team sends three experts to assess the science and technology talents under each indicator, and 12 experts from each of the four teams fill in the assessment values under the percentage system, and this paper uses the interval grey number to characterize the assessment values of the new R&D organization teams on science and technology talents, which indicates the uncertainty of the assessment values under this indicator.

The data in the above table do not reveal the overall situation, the comparison of strengths and weaknesses, and the proposed matching characteristics of the assessment of scientific and technological talents S by the new R&D organization team T. Therefore, certain methods are needed to transform and assemble the data so that the information of the data can reflect the characteristics of the assessment object.

In this paper, based on all the assessed values of the j-th indicator under a ¯ of the maximum value is normalized to obtain the normative matrix. The correlation coefficient of each assessment value with the maximum assessment sequence is calculated based on the grey correlation analysis formula, the larger the correlation coefficient, the higher the satisfaction of the new R&D organization team T with the scientific and technological talents S under this indicator. The correlation coefficient of T-S is shown in Figs. 3–8.

OPEN IN VIEWER

Fig. 3

Satisfaction of new R&D organization teams T with scientific and technological talent S1.

OPEN IN VIEWER

Fig. 4

Satisfaction of new R&D organization teams T with scientific and technological talent S2.

OPEN IN VIEWER

Fig. 5

Satisfaction of new R&D organization teams T with scientific and technological talent S3.

OPEN IN VIEWER

Fig. 6

Satisfaction of new R&D organization teams T with scientific and technological talent S4.

OPEN IN VIEWER

Fig. 7

Satisfaction of new R&D organization teams T with scientific and technological talent S5.

OPEN IN VIEWER

Fig. 8

Satisfaction of new R&D organization teams T with scientific and technological talent S6.

As can be seen from the above graph: team T1 rated talent S1 higher on the first 5 indicators and lower on the last 5 indicators; team T2 kept the rating of talent S1 stable and rated higher under the last 2 indicators, considering S1’s technological ability to be stronger; team T3 was more concerned about talent S1 being a graduate of a prestigious university and having strong technological insight; team T4 considered talent S1 to be more self-reliant and self-sufficient.

As can be seen from the above graph: team T1 rated talent S2 low; team T2 focused on talent S2’s innovative achievements and cross-disciplinary background; team T3 thought talent S2 had more innovative achievements and graduated from a prestigious university; team T4 thought talent S2 had a strong will to serve the country and was able to exercise self-control and self-discipline.

As can be seen from the above graph: team T1 thinks that talent S3 has a strong will to serve the motherland, has fruitful innovation results, and values more the fact that it graduated from a prestigious university; team T2 has a low overall rating for talent S2, especially that S2 cannot stand on its own feet; team T3 is concerned about talent S3’s self-awareness and self-confidence to engage in science and technology; team T4 thinks that talent S3 has more scientific and technological cooperation.

As can be seen from the above graph: team T1 perceives talent S4 to be scientifically and technologically knowledgeable and more technologically competent; team T2 has a low overall rating for talent S4; team T3 is concerned that talent S4 also has a low overall rating, particularly perceiving it to be less capable of scientific and technical collaboration; team T4 has a high rating for talent S4, perceiving S4 to be more entrepreneurial and to have a cross-disciplinary background, while being able to exercise self-control and self-discipline.

As can be seen from the above graph: team T1 thinks that talent S5 is self-aware and confident; team T2 also thinks that talent S5 is self-aware and confident, but the ability to cooperate in science and technology needs to be strengthened; team T3 thinks that talent S5 has more innovative achievements and graduated from a prestigious university; team T4 rates talent S5 higher and thinks that S5 has a strong ability to cooperate in science and technology and has a cross-disciplinary background, while being able to be self-sufficient and self-reliant.

As can be seen from the chart above: team T1 believes that talent S6 has more entrepreneurial achievements and is more self-controlled and self-disciplined; team T2 also believes that talent S6 has more entrepreneurial achievements; team T3 believes that talent S6 has a strong will to serve his country; team T4 believes that talent S6 is self-aware and confident, graduated from a prestigious school and is self-sufficient.

In this paper, 12 experts were consulted to obtain the subjective weights of the indicators ω_1j. The objective weights of the indicators are obtained based on the share of the sum of the correlation coefficients ω_2j, introduce the parameter adjustment parameter θ = 0.5, the composite indicator weights are obtained ω_3j, as shown in Fig. 9.

OPEN IN VIEWER

Fig. 9

Subjective-objective-composite indicator weights for team assessment of talent.

From the above Fig., it can be seen that: the T-S subjective weight is more biased towards innovation results and entrepreneurship results, and gives lower weight to serving the motherland, conscious self-confidence, graduation from prestigious schools and scientific and technological cooperation, while the T-S objective weight has little difference and is biased towards serving the motherland, conscious self-confidence, innovation results and graduation from prestigious schools, etc. The T-S combined weight is more reasonable between the subjective and objective weights.

According to the integrated weights, the correlation coefficient based on the weights is calculated and the correlation coefficient matrix under the weights is obtained. According to the 3E cloud model inverse generator operation rules, the cloud data of the new R&D organization team’s assessment of scientific and technological talents are obtained, as shown in Table 10 and Figs. 9–14.

OPEN IN VIEWER

Fig. 10

Cloud model diagram for S1 assessment by team T in new R&D organizations.

OPEN IN VIEWER

Fig. 11

Cloud model diagram for S2 assessment by team T in new R&D organizations.

OPEN IN VIEWER

Fig. 12

Cloud model diagram for S3 assessment by team T in new R&D organizations.

OPEN IN VIEWER

Fig. 13

Cloud model diagram for S4 assessment by team T in new R&D organizations.

OPEN IN VIEWER

Fig. 14

Cloud model diagram for S5 assessment by team T in new R&D organizations.

The cloud data in the table is not intuitive, and in this paper, the weight-based correlation coefficients are input into MATLAB for the simulation of the cloud model, and the cloud model diagram of the assessment of S for scientific and technological talents by the team T of the new R&D organization is obtained, as shown in Figs. 9–15.

OPEN IN VIEWER

Fig. 15

Cloud model diagram for S6 assessment by team T in new R&D organizations.

As can be seen from the graph above: teams T1-T4 all have assessment expectations around 0.5 for tech talent S1, and team T3 has a higher degree of cohesive cloud-drop hugging.

As can be seen from the above graph: team T1-T4 all have less than 0.5 assessment expectations for tech talent S2, and team T2 has a higher degree of cloud-drop cohesive hugging.

As can be seen from the above graph: teams T1-T4 all have assessment expectations around 0.4 for tech talent S3, with team T1 having a higher level of assessment cloud drops cohesive hugging.

As can be seen from the above graph: none of the team T1-T3’s assessment expectations for tech talent S4 exceed 0.5, team T4’s assessment expectations for tech talent S3 exceed 0.5, and team T4’s assessment cloud drops a higher degree of cohesive hugging.

As can be seen from the above graph: team T1-T3 all have assessment expectations around 0.4 for tech talent S5, team T4 has assessment expectations over 0.5 for tech talent S3, and team T1 has a higher level of assessment cloud drops cohesive hugging.

As can be seen from the above graph: the assessment expectations of Team T1-T4 for S5 of tech talent are all around 0.4, and Team T4 has a higher level of cohesive cloud drops hugging.

3.3.1 Collect assessment data and characterize the initial assessment values in the form of a standard cloud

In this paper, a questionnaire was sent out to the scientific and technical talents who participated in the matching team, and the scientific and technical talents were asked to check the measures under the assessment index directly on the questionnaire, which were divided into 5 levels (Very poor, Poor, Fair, Good and Excellent). In the process of returning the questionnaire, it was found that some scientific and technological talents did not limit the results to the 5 levels of measures, for example, S1 checked “poor” and “Fair”, and the assessment value of talent S1 was considered to be between “poor” and “Fair”. This paper considers that the assessment value of talent S1 is between “poor” and “Fair”, so the middle value is taken, which is simple to deal with, and the middle value is more representative, taking into account the two assessment measures. The assessment of scientific and technological talents S on the new R&D organization team T is shown in Tables 11–16.

The language assessment values from Tables 11–16 were found according to the standard cloud to find the expectation values, as shown in Table 17.

The data in the above table do not reveal the overall situation, the comparison of strengths and weaknesses, and the proposed matching characteristics of the assessment of the new R&D organization team T by the scientific and technological talent S. Therefore, certain methods are needed to transform and assemble the data so that the information of the data can reflect the characteristics of the assessment object.

This paper calculates the correlation coefficient of each assessment value with the maximum assessment sequence based on the grey correlation analysis formula, the larger the correlation coefficient, the higher the satisfaction of the scientific and technological talents S with the new R&D organization team T in this index. The correlation coefficient of S-T is shown in Figs. 16–21.

OPEN IN VIEWER

Fig. 16

Satisfaction of scientific and technological talents S1 with the new R&D organization teams T.

OPEN IN VIEWER

Fig. 17

Satisfaction of scientific and technological talents S2 with the new R&D organization teams T.

OPEN IN VIEWER

Fig. 18

Satisfaction of scientific and technological talents S3 with the new R&D organization teams T.

OPEN IN VIEWER

Fig. 19

Satisfaction of scientific and technological talents S4 with the new R&D organization teams T.

OPEN IN VIEWER

Fig. 20

Satisfaction of scientific and technological talents S5 with the new R&D organization teams T.

OPEN IN VIEWER

Fig. 21

Satisfaction of scientific and technological talents S6 with the new R&D organization teams T.

As can be seen from the chart above: talent S1 is satisfied with team T1’s talent ladder and talent treatment; satisfied with team T2’s talent training and talent motivation and dissatisfied with its talent ladder and talent treatment; satisfied with team T3’s talent treatment and dissatisfied with its talent ladder; and rated team T4 relatively low.

As you can see from the chart above: talent S2 is satisfied with the talent treatment of team T1 and very dissatisfied with their talent incentives; they are satisfied with the talent ladder of team T2 and very dissatisfied with their talent treatment; they are satisfied with the talent training and talent treatment of team T3; and they are satisfied with the talent treatment of team T4.

As you can see from the chart above: talent S3 is very satisfied with the talent training and talent treatment of team T1 and very dissatisfied with the talent ladder; very satisfied with the talent ladder of team T2; very satisfied with the talent treatment and talent motivation of team T3; and rated team T4 relatively low.

The above chart shows that: talent S4 is satisfied with the talent treatment of team T1; dissatisfied with all indicators of team T2; satisfied with the talent training, talent ladder, and talent motivation of team T3 and dissatisfied with its talent treatment; satisfied with the talent training of team T4, but dissatisfied with other indicators.

As can be seen from the chart above: talent S5 is satisfied with the talent treatment of team T1; satisfied with the talent training of team T2 and dissatisfied with its other indicators; satisfied with the talent training and talent motivation of team T3 and very dissatisfied with its talent ladder; satisfied with the talent training of team T4 and dissatisfied with its other indicators.

The chart above shows that Talent S6 is satisfied with Team T1’s talent treatment; satisfied with Team T2’s talent training and talent ladder; rated Team T3 poorly on all indicators; satisfied with Team T4’s talent training and talent motivation, and dissatisfied with its talent ladder and talent treatment.

In this paper, some scientific and technical talents were consulted to obtain the subjective weights of the indicators ω_4j. The objective weights of the indicators are obtained based on the share of the sum of the correlation coefficients ω_5j, introduce the parameter adjustment parameter δ = 0.5, the comprehensive indicator weights are obtained ω_6j, as shown in Fig. 22.

OPEN IN VIEWER

Fig. 22

Subjective-objective-composite indicator weights for talent-to-team assessment.

As can be seen from the chart above: the subjective weight is more biased towards talent treatment, giving a lower weight to talent treatment, while the objective weight is less disparate and biased towards talent training and talent treatment, and the combined weight is more reasonable between the subjective and objective weights.

According to the integrated weights, the correlation coefficient based on the weights is calculated and the correlation coefficient matrix under the weights is obtained. According to the 3E cloud model inverse generator operation rules, the cloud data of the evaluation of scientific and technological talents to the team of new R&D organization is obtained, as shown in Table 18 and Fig. 23–28. In order to facilitate the bilateral matching later, the cloud data are transposed in this paper.

OPEN IN VIEWER

Fig. 23

Cloud model diagram for the assessment of team T in new R&D institutions by S1.

OPEN IN VIEWER

Fig. 24

Cloud model diagram for the assessment of team T in new R&D institutions by S2.

OPEN IN VIEWER

Fig. 25

Cloud model diagram for the assessment of team T in new R&D institutions by S3.

OPEN IN VIEWER

Fig. 26

Cloud model diagram for the assessment of team T in new R&D institutions by S4.

OPEN IN VIEWER

Fig. 27

Cloud model diagram for the assessment of team T in new R&D institutions by S5.

OPEN IN VIEWER

Fig. 28

Cloud model diagram for the assessment of team T in new R&D institutions by S6.

The cloud data in the table is not intuitive, and in this paper, the weight-based correlation coefficients are input into MATLAB for the simulation of the cloud model, and the cloud model diagram of the assessment of S for scientific and technological talents by the team T of the new R&D organization is obtained, as shown in Figs. 23–28.

As can be seen from the above graph: the assessment expectation of tech talent S1 for team T1 exceeds 0.5, the assessment expectation of team T2, T3 and T4 do not exceed 0.5, and the assessment of team T2 has a higher degree of cloud-drop cohesive hugging.

As can be seen from the above graph: the assessment expectations of S2 for tech talent for team T1, T3, and T4 are more than 0.5, the assessment expectations for team T2 are less than 0.5, and the assessment of team T3 has a higher degree of cloud-drop cohesive hugging.

As can be seen from the above graph: the assessment expectations of the tech talent S3 for team T1, T2, and T3 are more than 0.5, the assessment expectations for team T2 are less than 0.5, and the assessment of team T2 has a higher degree of cloud-drop cohesive hugging.

As can be seen from the above graph: the assessment expectations of tech talent S4 for team T1 and T3 are more than 0.5, the assessment expectations for team T2 and T4 are less than 0.5, and the assessment of team T3 has a higher degree of cloud-drop cohesive hugging.

As can be seen from the above graph: the assessment expectations of S5 of tech talent for team T1, T3 and T4 are more than 0.5 and less than 0.5 for team T2, their assessment cloud drops cohesive hugging is about the same and all are more discrete.

As can be seen from the above graph: the assessment expectation of S5 for tech talent is more than 0.5 for team T1 and T2, around 0.5 for team T4, less than 0.5 for team T3, and a higher degree of cloud-drop cohesive hugging for team T3.