Regional standardized development data comparison method and system based on quantitative model

Abstract

The invention discloses a regional standardized development data comparison method and system based on a quantitative model, and the method comprises the steps: obtaining regional standardized development data, carrying out the preprocessing of the development data, carrying out the feature screening of historical feature factors, obtaining a feature subset, processing the feature subset, obtaining an index feature library, selecting to-be-compared indexes according to the correlation degree to obtain key indexes, matching the key indexes with standard indexes of the index feature library to obtain to-be-compared index groups, and constructing a neural network-based regional standardized development data quantification model according to the index feature library, and inputting the to-be-compared index group into the regional standardized development data quantification model, and outputting a comparison result of the to-be-compared indexes and the index feature library. The method not only can improve the precision and speed of regional standardized development data comparison, but also has good interpretability, and can be directly applied to a regional standardized development data comparison system.

Description

Technical Field

[0001] The present invention relates to the field of data comparison, and in particular to a method and system for comparing regional standardized development data based on a quantitative model. Background Art

[0002] Regional standardization refers to standardization activities carried out by cities or countries in the same region. It is an important means to promote trade facilitation, technological interoperability and economic development between the region or adjacent regions. Quantitative models are mainly based on historical data and use mathematical and statistical methods to analyze and predict data behavior.

[0003] Regional standardization plays an important role in the process of global integration. The standardization development data of different regions are diverse, complex and heterogeneous, and it is difficult to directly compare and analyze these data. The current regional standardization development data comparison technology has problems such as insufficient data preprocessing ability, single model construction, low comparison efficiency and poor result interpretation. Combining quantitative models can assist decision makers and standardization workers to accurately match key indicators for data comparison and quantitatively evaluate standardization development data in complex and changing regional environments, thereby achieving the optimization formulation and effective implementation of regional standards. Therefore, designing a regional standardization development data comparison method based on quantitative models to overcome the shortcomings of existing comparison methods is of great significance for the in-depth promotion of regional standardization work. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for comparing regional standardized development data based on a quantitative model.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Acquire regional standardized development data and pre-process the development data; the development data includes historical indicators and indicators to be compared;

[0008] Extracting the characteristics of the historical indicator and the indicator to be compared to obtain a historical characteristic factor and a comparison characteristic factor, performing feature screening on the historical characteristic factor to obtain a feature subset, and processing the feature subset to obtain an indicator feature library; the feature screening includes a first feature screening and a second feature screening; the feature subset includes a first feature subset and a second feature subset;

[0009] Selecting the indicators to be compared according to the correlation to obtain key indicators, and matching the key indicators with the standard indicators in the indicator feature library to obtain the indicator group to be compared;

[0010] Construct a quantization model of regional standardized development data based on a neural network according to the index feature library, and input the index group to be compared into the quantization model of regional standardized development data to output the comparison result between the index to be compared and the index feature library.

[0011] Furthermore, the method for preprocessing the development data includes:

[0012] Clean, deduplicate, and normalize the development data, classify the development data using a decision tree algorithm, and perform a labeling operation on the development data according to the classification results.

[0013] Furthermore, the method for the first feature screening includes:

[0014] Use linear discriminant analysis feature extraction for historical indicators and indicators to be compared to obtain historical feature factors and comparison feature factors;

[0015] Define t ij as the j-dimensional feature corresponding to the i-th historical indicator, 1 < i < m, 1 < j < n, m is the number of historical indicators, n is the number of historical feature factors, and use the improved Harris hawk optimization algorithm to perform feature screening on the i-th historical indicator to obtain the first feature to form the first feature subset;

[0016] The improved Harris hawk optimization algorithm specifically includes:

[0017] Perform binary encoding on historical feature factors, and the values of historical feature factors are taken as 1 and 0 according to being selected and not being selected, and calculate the individual fitness of the hawk:

[0018]

[0019] where Fitness is the individual fitness of the hawk, (1 - Accurcy) is the classification error rate, N is the feature length, ∝ is the classification accuracy, and β = 1 - ∝ is the weight of the selected feature in the fitness function;

[0020] Introduce a chaotic map to update the rabbit escape energy formula:

[0021]

[0022] where E k+1 is the escape energy updated by the rabbit, E0 is the initial escape energy of the rabbit, E k = 2E0(1 - t / T) is the escape energy of the rabbit, t is the current iteration number, and T is the total number of iterations;

[0023] According to the updated rabbit escape energy E k+1 and the position update rule of the hawk, obtain the position update X(t + 1) of the hawk;

[0024] The eagle's position update rule is:

[0025] In the global search phase, the eagle's position update is expressed as:

[0026]

[0027] Where X(t) is the current position of the eagle, X(t+1) is the updated position of the eagle, t is the number of iterations, and X rabbit is the position of the rabbit, the average position of the eagle N is the total number of eagles, r1, r2, r3, r4, q are random numbers (0, 1), U B and L B are the upper and lower bounds of the variable, q is the eagle’s habitat pattern;

[0028] In the local search phase, four different hunting modes are carried out according to the rabbit's state. Let r be the chance of the rabbit escaping before the raid. If r < 0.5, there is a possibility of successful escape, and if r ≥ 0.5, there is no successful escape. Combined with the updated rabbit escape energy E k+1 , when r ≥ 0.5 and |E k+1 When |≥0.5, the eagle group adopts strategy 1 to attack, and the position update is expressed as:

[0029] X(t+1)=X rabbit (t)-X(t)-E k+1 |JX rabbit (t)-X(t)|

[0030] Among them, is the value representing the rabbit's motion properties in each iteration;

[0031] When r ≥ 0.5 and |E k+1 When |<0.5, the eagle group adopts strategy 2 to attack, and the position update is expressed as:

[0032] X(t+1)=X rabbit (t)-E k+1 |X rabbit (t)-X(t)|

[0033] When r < 0.5 and |E k+1 When |≥0.5, the eagle group adopts strategy 3 to attack, and the position update is expressed as:

[0034]

[0035] Where S is the value of an n-dimensional random vector element between [0, 1], Γ(·) is the probability density function, μ and v are random numbers between (0, 1), and β is the "jump" characteristic parameter, which is 1.5;

[0036] When r < 0.5 and |E k+1When |<0.5, the eagle group adopts strategy 4 to attack, and the position update is expressed as:

[0037]

[0038] Add normal distribution random numbers to perform Gaussian mutation on the eagle's position update:

[0039]

[0040] Where X(t+2) is the updated position of the eagle after Gaussian mutation, X best (t+2)=X(t+1)[1+N(μ,σ 2 )] is the value obtained by Gaussian mutation of the eagle's position X(t), N is a normally distributed random number, μ is the mean, σ 2 is the variance, f(·) is the fitness function, which indicates the fitness of the current position, rand is a random number, and p is the probability of survival of the fittest;

[0041] KNN is used to establish a parcel feature selection model to evaluate the quality of feature subsets, where k=5. The individual fitness of the eagle is continued to be calculated and the position of the eagle group is updated. The first feature is output after the maximum number of iterations is reached, and the first features of m historical indicators are combined to obtain the first feature subset.

[0042] Furthermore, the method for screening the second feature includes:

[0043] Define the historical indicator set E = {e1, e2, ..., e m}, m is the number of historical indicators, the i-th historical indicator e i Corresponding to the historical characteristic factor set T i ={t i1 , t i2 ,…,t in}, n is the number of historical characteristic factors;

[0044] e i Corresponding sample, t ij Corresponding sample e i The j feature of the sample center circle with a radius of r is divided into a parts, and the multi-directional neighbor search method is used to find the k nearest hit points H of the same category. j and k different categories of recent miss points H j , the weights are calculated using the feature weight objective equation:

[0045]

[0046] Where W(t ij ) is the characteristic t ij weight, C is the category, Y is the category label set, Class(ei ) is the sample e i Get category, y(e i ) is the instance label, P(C) is the prior probability of category C, c is the number of categories, n is the number of features, diff(A, B, C) is the difference function, which indicates the difference between A in B and C, M l (Class(e i )) is the closest miss in category C, H l is the closest hit of the same type, u i is the smaller weight coefficient given to the sample when the similarity between the sample and the farthest miss is large, h i A larger weight coefficient is given to the sample when the similarity between the sample and the farthest hit is large;

[0047] Filter feature selection by maximum Pierce correlation coefficient and maximum correlation distance:

[0048]

[0049] Where MPMD(t ij ) is t ij The feature weight of , t is the current iteration number, T is the total iteration number, a is a constant, PE(·) is the Pearson correlation coefficient, and the calculation rule is

[0050] Calculate feature t ij Probability of being selected:

[0051]

[0052] Among them, PP(t ij ) is the characteristic t ij The probability of being selected, β is MPMD(t ij ) and W(t ij ), where β=1 is assumed to be equally important. The historical characteristic factors are selected according to the probability of being selected, and the second characteristics of m historical indicators are selected to obtain the second characteristic subset.

[0053] Furthermore, the method of processing the feature subset to obtain an indicator feature library includes:

[0054] The intersection of the first feature subset and the second feature subset is taken as the feature representative, the unintersected parts S1 and S2 of the first feature subset and the second feature subset are taken to form the pending feature representative, the pending feature representative is screened twice to obtain the remaining feature representatives, and all feature representatives and corresponding historical indicators are combined to obtain the indicator feature library;

[0055] Methods for secondary screening of candidate feature representatives include:

[0056] Define the set \(T\) of undetermined characteristic representatives i =\(\{t\) i1 , \(t\) i2 , \(\cdots\), \(t\) ij , \(\cdots\), \(t\) il \}\), where \(t\) ij is the \(j\)-dimensional characteristic corresponding to the \(i\)-th index, \(1 \lt i \lt m\), \(1 \lt j \lt l\), \(m\) is the number of indices, and \(l\) is the number of non-intersecting historical characteristic factors. Randomly initialize the assignment of \(t\) ij :

[0057]

[0058] where \(rand\in(0, 1)\) is a random value;

[0059] Use the bird swarm algorithm to perform secondary screening on historical characteristic factors. The update formulas for the bird swarm velocity and position are:

[0060]

[0061] x ij (t + 1)=x ij (t)+v ij (t + 1)

[0062] where and \(x\) ij (t + 1) are the velocity and position of the bird at the \((t + 1)\)-th iteration, \(\omega\) is the inertia weight, \(c_1\) and \(c_2\) are learning factors, \(rand1, rand2\in(0, 1)\), \(pbest\) ij and \(gbest\) represent the optimal positions of the bird and the bird swarm respectively;

[0063] Perturb the index, and the expression is:

[0064]

[0065] Introduce the Dunn index Combine the stability measurement criterion to calculate the fitness:

[0066]

[0067] where \(Fitness\) is the fitness, \(0.33\) is the coefficient weight, \(errorrate\) is the classification error rate, \(dimension = l\) is the feature dimension, \(n\) is the total number of original features, \(min\delta(S1, S2)\) is the minimum value of the distance between historical characteristic factors of two non-intersecting sets, is the average value of the distance between historical characteristic factors of two non-intersecting sets;

[0068] According to the fitness calculation results, the undetermined feature representatives are screened to obtain the remaining feature representatives.

[0069] Furthermore, the method of selecting the indicators to be compared according to the correlation to obtain the key indicators includes:

[0070] Set the index to be compared i as the reference index X in turn i (j), i∈(0, m), m is the number of indicators to be compared, and the remaining indicators to be compared are set as the control indicators Y i (j), j∈(0,n) is the characteristic factor number of comparison index i, and n is the number of comparison characteristic factors;

[0071] Calculate the relationship number between the reference indicator and the control indicator:

[0072]

[0073] where d i (k) is the reference index X i (j) and control index Y i (j) In the correlation coefficient of the comparison characteristic factor j, ρ is the resolution coefficient and is taken as 0.5;

[0074] According to the weight w of the comparison feature factor j Calculate the correlation score between the reference indicator and the control indicator:

[0075]

[0076] Where R i X is the reference index i (j) and control index Y i (j) total relevance score, where a is the relevance score coefficient, and taking a = 10 changes the relevance score to a 10-point scale;

[0077] The correlation of the indicators to be compared is evaluated by comprehensive grey correlation scores, and the indicators to be compared with high correlation are selected to obtain key indicators.

[0078] Furthermore, the method of matching the key indicator with the standard indicator of the indicator feature library to obtain the indicator group to be compared includes:

[0079] Rough matching: defining key indicator A i =(a i1 , a i2 , …, a ix , …, a im ) and standard indicator B in the indicator feature library j =(b j1 , b j2 , …, b jy , …, b jn), calculate key indicator A according to the indicator classification i All standard indicators of the same category B j The cosine value of is used to identify similar characteristic factor combinations according to the cosine value to obtain a rough matching index group;

[0080] The key indicator A i The formula for calculating the cosine value of all standard indicators B in the same category is:

[0081]

[0082] Where cos(A i , B j ) is the key indicator A i With standard indicator B j Cosine value of eigenvector, key indicator A i The xth key feature factor vector is a ix , the number of eigenvectors is m, the standard index B j The yth standard eigenfactor vector of jy , the number of eigenvectors is n;

[0083] According to the calculation results, retain the key indicators A i 5 standard indicators with the largest cosine values ​​B j ;

[0084] Fine matching: focus on indicator A i And the corresponding standard indicators Feature alignment obtains feature factor vectors with the same dimension and calculates the key index A i With standard indicators The weighted Mahalanobis distance is calculated based on the weighted Mahalanobis distance, and the key indicator A is matched according to the weighted Mahalanobis distance calculation result. i With standard indicators

[0085] Key Indicator A i The key characteristic factor vector is a i =(a i1 , a i2 , …, a ix , …, a in ), 1≤i≤m, m is the number of key indicators, n is the length of the feature vector, corresponding to the standard indicator The standard eigenfactor vector is j is an integer between 1 and 5;

[0086] The subjective weight is calculated by the hierarchical analysis method, and the expression is:

[0087]

[0088] where wz Standard eigenfactors calculated for the AHP The subjective weight of is the weight calculated by normalizing using the arithmetic mean method, is the weight calculated by geometric mean normalization, w z3 is the weight calculated by the eigenvalue method;

[0089] The objective weight is calculated by the CRITIC method, and the expression is:

[0090]

[0091] where w K Standard characteristic factors calculated by the CRITIC method The objective weight of j is the comparison strength, R j It is a conflict indicator;

[0092] The comprehensive weight is obtained based on the subjective weight and the objective weight, and the expression is:

[0093] w ij =τw z (ij)+vw K (ij)

[0094] where w ij is the standard characteristic factor The comprehensive weight of , τ and v are weight coefficients, τ+v=1;

[0095] Mean vector The covariance is Key Indicator A i With standard indicators The weighted Mahalanobis distance calculation is expressed as:

[0096]

[0097] Take the key index A corresponding to the minimum weighted Mahalanobis distance i and standard indicators is a pair of matching indicators to be compared, and m pairs of matching indicators to be compared are formed into an indicator group to be compared.

[0098] Furthermore, a method for constructing a regional standardized development data quantification model based on a neural network according to the indicator feature library includes:

[0099] The standard indicators, standard indicator categories, and feature representatives of the standard indicators in the indicator feature library are combined into a standard vector set, and the standard vector set is divided into a training set and a test set;

[0100] Input the training set into the XGBoost enhanced BP neural network for training;

[0101] Use the XGBoost model to predict the input training set, output new features and add them to the original training set, input the training set into the BP neural network for training, select the cross entropy loss function to evaluate the model, and select the SGD optimizer based on the evaluation to update the network parameters and output the development data quantization function;

[0102] The test set is used to evaluate the development data quantification function and build a regional standardized development data quantification model:

[0103]

[0104] Where S i is the quantitative model comparison score of the i-th indicator, S is the quantitative model comparison score of the user as a whole, and n i is the number of features of the ith indicator, m is the number of indicators of the user, is the category weight of the i-th indicator, is the indicator weight of the i-th indicator, I ij is the score represented by the jth feature of the i-th indicator;

[0105] The indicator group to be compared is input into the regional standardized development data quantification model to output the comparison results between the indicator to be compared and the indicator feature library.

[0106] Second, a regional standardized development data comparison system based on a quantitative model, including:

[0107] Data collection module: used to collect regional standardized development data, pre-process the development data, and transmit the development data to the screening module and the matching module; the development data includes historical indicators and indicators to be compared;

[0108] Screening module: used to screen the historical characteristic factors to obtain an indicator characteristic library;

[0109] Matching module: used for selecting the indicators to be compared according to the correlation to obtain the key indicators, and matching the key indicators with the standard indicators in the indicator feature library to obtain the indicator group to be compared;

[0110] Machine learning module: used for inputting the standard feature library into the XGBoost enhanced BP neural network to obtain a regional standardized development data quantification model;

[0111] Comparison module: used to manage the development data, the indicator feature library and the indicator group to be compared, input the indicator group to be compared into the regional standardized development data quantification model, and output the comparison result between the indicator to be compared and the indicator feature library.

[0112] The beneficial effects of the present invention are:

[0113] The present invention is a method and system for comparing regional standardized development data based on a quantitative model. Compared with the prior art, the present invention has the following technical effects:

[0114] The present invention can improve the data preprocessing capabilities and enhance the model adaptability in the regional standardized development data comparison by extracting indicator features, screening features, obtaining an indicator feature library, obtaining key indicators, obtaining a group of indicators to be compared, and constructing a model, thereby improving the comparison efficiency and accuracy of the regional standardized development data comparison, and optimizing the regional standardized development data comparison technology, which can greatly save resources and improve work efficiency. It can realize the comparison of regional standardized development data and provide strong support for the formulation of decisions and plans. It is of great significance to the regional standardized development data comparison, and can adapt to different comparison systems of regional standardized development data based on quantitative models and terminal data comparison needs of comparison systems of regional standardized development data based on quantitative models of different users, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 The present invention is a flowchart of the steps of a method for comparing regional standardized development data based on a quantitative model. DETAILED DESCRIPTION

[0116] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0117] The present invention provides a method and system for comparing regional standardized development data based on a quantitative model, comprising the following steps:

[0118] like Figure 1 As shown, in this embodiment, the following steps are included:

[0119] Acquire regional standardized development data and pre-process the development data; the development data includes historical indicators and indicators to be compared;

[0120] Extracting the characteristics of the historical indicator and the indicator to be compared to obtain a historical characteristic factor and a comparison characteristic factor, performing feature screening on the historical characteristic factor to obtain a feature subset, and processing the feature subset to obtain an indicator feature library; the feature screening includes a first feature screening and a second feature screening; the feature subset includes a first feature subset and a second feature subset;

[0121] Selecting the indicators to be compared according to the correlation to obtain key indicators, and matching the key indicators with the standard indicators in the indicator feature library to obtain the indicator group to be compared;

[0122] Construct a neural network-based regional standardization development data quantization model according to the index feature library, and input the index group to be compared into the regional standardization development data quantization model to output the comparison result between the index to be compared and the index feature library.

[0123] In this embodiment, the method for preprocessing the development data includes:

[0124] Clean, de-duplicate and normalize the development data, classify the development data using the decision tree algorithm, and perform label operation on the development data according to the classification result;

[0125] In actual evaluation, taking the standardization development situation of Region X as the research object, a set of historical indicators and classification situations are given: (Regional economic development level category: Indicator 1. Development level / Good, 2. Fixed asset investment / Excellent), (Regional integration level category: Indicator 1. Market integration / Excellent, 2. Passenger volume / Good, 3. Transportation integration / Excellent), (Urban-rural coordinated development level category: Indicator 1. Urban-rural income / Medium, 2. Urbanization / Good), (Social coordinated development level category: Indicator 1. Assistance situation / Excellent, 2. Education level / Good), (Resource and environment coordinated development level category: Indicator 1. Energy consumption / Excellent, 2. Pollution control / Good, 3. Ecological construction / Excellent);

[0126] A set of indicators to be compared and classification situations are given: (Regional economic development level category: Indicator 1. Innovation input / Excellent, 2. Industrial structure / Excellent, Foreign trade / Medium), (Regional integration level category: Indicator 1. Regional cooperation / Excellent, 2. Transportation integration / Good, 3. Information integration / Excellent), (Urban-rural coordinated development level category: Indicator 1. Urban-rural public services / Good, 2. Urban-rural environment / Excellent), (Social coordinated development level category: Indicator 1. Social security system / Excellent, 2. Public security / Good), (Resource and environment coordinated development level category: Indicator 1. Resource utilization efficiency / Excellent, 2. Ecological protection / Good, 3. Environmental governance / Excellent).

[0127] In this embodiment, the method for the first feature screening includes:

[0128] Use linear discriminant analysis feature extraction for historical indicators and indicators to be compared to obtain historical feature factors and comparison feature factors;

[0129] Define t ij as the j-dimensional feature corresponding to the i-th historical indicator, 1 < i < m, 1 < j < n, m is the number of historical indicators, n is the number of historical feature factors, and use the improved Harris hawk optimization algorithm to perform feature screening on the i-th historical indicator to obtain the first feature to form the first feature subset;

[0130] The improved Harris hawk optimization algorithm specifically includes:

[0131] The historical characteristic factors are binary-coded, and the values ​​of the historical characteristic factors are 1 and 0 according to whether they are selected or not, and the fitness of the eagle individuals is calculated:

[0132]

[0133] Where Fitness is the fitness of the eagle individual, (1-Accurcy) is the classification error rate, N is the feature length, ∝ is the accuracy of classification, and β=1-∝ is the weight of the selected feature in the fitness function;

[0134] Introduce chaos mapping to update the rabbit escape energy formula:

[0135]

[0136] Where E k+1 is the updated escape energy of the rabbit, E0 is the initial energy of the rabbit’s escape, E k =2E0(1-t / T) is the rabbit's escape energy, t is the current iteration number, and T is the total iteration number;

[0137] According to the updated bunny escape energy E k+1 And the Eagle's position update rule obtains the Eagle's position update X(t+1);

[0138] The eagle's position update rule is:

[0139] In the global search phase, the eagle's position update is expressed as:

[0140]

[0141] Where X(t) is the current position of the eagle, X(t+1) is the updated position of the eagle, t is the number of iterations, and X rabbit is the position of the rabbit, the average position of the eagle N is the total number of eagles, r1, r2, r3, r4, q are random numbers (0, 1), U B and L B are the upper and lower bounds of the variable, q is the eagle’s habitat pattern;

[0142] In the local search phase, four different hunting modes are carried out according to the rabbit's state. Let r be the chance of the rabbit escaping before the raid. If r < 0.5, there is a possibility of successful escape, and if r ≥ 0.5, there is no successful escape. Combined with the updated rabbit escape energy E k+1 , when r ≥ 0.5 and |E k+1 When |≥0.5, the eagle group adopts strategy 1 to attack, and the position update is expressed as:

[0143] X(t+1)=X rabbit(t)-X(t)-E k+1 |JX rabbit (t)-X(t)|

[0144] Among them, is the value representing the rabbit's motion properties in each iteration;

[0145] When r ≥ 0.5 and |E k+1 When |<0.5, the eagle group adopts strategy 2 to attack, and the position update is expressed as:

[0146] X(t+1)=X rabbit (t)-E k+1 |X rabbit (t)-X(t)|

[0147] When r<0.5 and |E|≥0.5, the eagle group adopts strategy 3 to attack, and the position update is expressed as:

[0148]

[0149] Where S is the value of an n-dimensional random vector element between [0, 1], Γ(·) is the probability density function, μ and v are random numbers between (0, 1), and β is the "jump" characteristic parameter, which is 1.5;

[0150] When r < 0.5 and |E k+1 When |<0.5, the eagle group adopts strategy 4 to attack, and the position update is expressed as:

[0151]

[0152] Add normal distribution random numbers to perform Gaussian mutation on the eagle's position update:

[0153]

[0154] Where X(t+2) is the updated position of the eagle after Gaussian mutation, X best (t+2)=X(t+1)[1+N(μ,σ 2 )] is the value of the eagle's position X(t) after Gaussian mutation, N is a normal distribution random number, μ is the mean, σ 2 is the variance, f(·) is the fitness function, which indicates the fitness of the current position, rand is a random number, and p is the probability of survival of the fittest;

[0155] KNN is used to establish a wrap-around feature selection model to evaluate the quality of feature subsets, where k = 5. The fitness of individual eagles and the position of the eagle group are continuously calculated, and the first feature is output after the maximum number of iterations is reached. The first features of m historical indicators are combined to obtain the first feature subset.

[0156] In the actual evaluation, taking the regional economic development level indicators as an example, the historical indicator characteristics were screened to obtain the historical characteristic factors: (Regional economic development level category: 1. Development level / good: annual GDP growth rate -5.2%, per capita GDP -$45,000, economic growth stability index -8.3; 2. Fixed asset investment / excellent: annual fixed asset investment amount -$20 billion, number of investment projects -150, investment benefit ratio -1.9, foreign investment attraction index -7.5)

[0157] Taking the regional economic development level indicators as an example, the comparison indicator characteristics are screened to obtain the comparison characteristic factors (regional economic development level: indicators 1. Innovation investment / excellent: the proportion of R&D expenditure to GDP-3.0%, the number of patent applications per million people-150, the density of scientific and technological innovation enterprises-120 enterprises / 10,000 people, the conversion rate of scientific and technological achievements-75%, the intensity of innovation capital investment-$500 / person; 2. Industrial structure / excellent: the proportion of the tertiary industry-70%, the proportion of high-tech industry added value-28%, the speed of industrial upgrading-4.5%, the industrial diversification index-8.2; 3. Foreign trade / medium: the growth rate of total exports-6.0%, the growth rate of total imports-5.5%, the trade surplus-$15 billion)

[0158] In this embodiment, the second feature screening method includes:

[0159] Define the historical indicator set E = {e1, e2, ..., e m}, m is the number of historical indicators, the i-th historical indicator e i Corresponding to the historical characteristic factor set T i ={t i1 , t i2 ,…,t in}, n is the number of historical characteristic factors;

[0160] e i Corresponding sample, t ij Corresponding sample e i The j feature of the sample center circle with a radius of r is divided into a parts, and the multi-directional neighbor search method is used to find the k nearest hit points H of the same category. j and k different categories of recent miss points H j , the weights are calculated using the feature weight objective equation:

[0161]

[0162] Where W(t ij ) is the characteristic t ij weight, C is the category, Y is the category label set, Class(e i ) is the sample ei Get category, y(e i ) is the instance label, P(C) is the prior probability of category C, c is the number of categories, n is the number of features, diff(A, B, C) is the difference function, which indicates the difference between A in B and C, M l (Class(e i )) is the closest miss in category C, H l is the closest hit of the same type, u i is the smaller weight coefficient given to the sample when the similarity between the sample and the farthest miss is large, h i A larger weight coefficient is given to the sample when the similarity between the sample and the farthest hit is large;

[0163] Filter feature selection by maximum Pierce correlation coefficient and maximum correlation distance:

[0164]

[0165] Where MPMD(t ij ) is t ij The feature weight of , t is the current iteration number, T is the total iteration number, a is a constant, PE(·) is the Pearson correlation coefficient, and the calculation rule is

[0166] Calculate feature t ij Probability of being selected:

[0167]

[0168] Among them, PP(t ij ) is the characteristic t ij The probability of being selected, β is MPMD(t ij ) and W(t ij ), where β=1 is assumed to be equally important. The historical characteristic factors are selected according to the probability of being selected, and the second characteristics of m historical indicators are selected to obtain the second characteristic subset;

[0169] In the actual evaluation, taking the regional economic development level indicators as an example, the results of the second feature screening are shown: (Regional economic development level category: 1. Development level / good: annual GDP growth rate -76, per capita GDP -38.89, economic growth volatility -50.2; 2. Fixed asset investment / excellent: annual fixed asset investment amount -33.33, investment benefit ratio -30, number of investment projects -40).

[0170] In this embodiment, the method of processing the feature subset to obtain an indicator feature library includes:

[0171] Take the intersection of the first feature subset and the second feature subset as the feature representative, and take the non - intersecting parts S1 and S2 of the first feature subset and the second feature subset to form the to - be - determined feature representative. Perform a secondary screening on the to - be - determined feature representative to obtain the remaining feature representative, and combine all the feature representatives and the corresponding historical indicators to obtain the index - feature library;

[0172] The method for performing a secondary screening on the to - be - determined feature representative includes:

[0173] Define the to - be - determined feature representative set T i ={t i1 , t i2 , …, t ij , …, t il}, where t ij is the j - dimensional feature corresponding to the i - th index, 1 < i < m, 1 < j < l, m is the number of indicators, and l is the number of non - intersecting historical feature factors. Perform a random initial assignment on t ij :

[0174]

[0175] where rand ∈ (0, 1) is a random value;

[0176] Use the bird - flock algorithm to perform a secondary screening on the historical feature factors. The bird - flock speed and position update formulas are:

[0177]

[0178] x ij (t + 1)=x ij (t)+v ij (t + 1)

[0179] where and x ij (t + 1) are the speed and position of the bird at the (t + 1) - th iteration, ω is the inertia weight, c1 and c2 are learning factors, rand1, rand2 ∈ (0, 1), pbest ij and gbest represent the optimal positions of the bird and the bird - flock respectively;

[0180] Perturb the indicators, and the expression is:

[0181]

[0182] Introduce the Dunn index Combine the stability metric criterion to calculate the fitness:

[0183]

[0184] Where Fitness is the fitness, 0.33 is the coefficient weight, errorrate is the classification error rate, dimension=l is the feature dimension, n is the original total number of features, minδ(S1, S2) is the minimum value of the historical feature factor distance between two disjoint sets, is the average value of the historical characteristic factor distance between two disjoint sets;

[0185] According to the fitness calculation results, the undetermined feature representatives are screened to obtain the remaining feature representatives;

[0186] In the actual evaluation, taking the regional economic development level indicators as an example, the results of the secondary screening are shown: (Regional economic development level: indicators 1. Development level / good: annual GDP growth rate, per capita GDP -38.89, economic growth volatility -50.2; 2. Fixed asset investment / excellent: annual fixed asset investment amount -33.33, investment benefit ratio -30).

[0187] In this embodiment, the method of selecting the indicators to be compared according to the correlation to obtain the key indicators includes:

[0188] Set the index i to be compared as the reference index X in turn i (j), i∈(0, m), m is the number of indicators to be compared, and the remaining indicators to be compared are set as the control indicators Y i (j), j∈(0,n) is the characteristic factor number of comparison index i, and n is the number of comparison characteristic factors;

[0189] Calculate the relationship number between the reference indicator and the control indicator:

[0190]

[0191] where d i (k) is the reference index X i (j) and control index Y i (j) In the correlation coefficient of the comparison characteristic factor j, ρ is the resolution coefficient and is taken as 0.5;

[0192] According to the weight w of the comparison feature factor j Calculate the correlation score between the reference indicator and the control indicator:

[0193]

[0194] Where R i X is the reference index i (j) and control index Y i (j) total relevance score, where a is the relevance score coefficient, and taking a = 10 changes the relevance score to a 10-point scale;

[0195] The correlation of the indicators to be compared is evaluated by comprehensive grey correlation scores, and the indicators to be compared with high correlation are selected to obtain key indicators;

[0196] In the actual evaluation, after the indicators to be compared are standardized, taking the regional economic development level indicators as an example, the indicators to be compared with higher correlation are selected to obtain the key indicators: (Regional economic development level category: Indicators 1. Innovation investment / excellent: R&D expenditure as a proportion of GDP -75, number of patent applications per million people -75, density of science and technology innovation enterprises -80, conversion rate of scientific and technological achievements -83, intensity of innovation capital investment -50; 2. Industrial structure / excellent: proportion of the tertiary industry -93, proportion of high-tech industry added value -93, speed of industrial upgrading -90, industrial diversification index -82).

[0197] In this embodiment, the method of matching the key indicator with the standard indicator of the indicator feature library to obtain the indicator group to be compared includes:

[0198] Rough matching: defining key indicator A i =(a i1 , a i2 , …, a ix , …, a im ) and standard indicator B in the indicator feature library j =(b j1 , b j2 , …, b jy , …, b jn ), calculate key indicator A according to the indicator classification i All standard indicators of the same category B j The cosine value of is used to identify similar characteristic factor combinations according to the cosine value to obtain a rough matching index group;

[0199] The key indicator A i All standard indicators of the same category B j The formula for calculating the cosine value of is:

[0200]

[0201] Where cos(A i , B j ) is the key indicator A i With standard indicator B j Cosine value of eigenvector, key indicator A i The xth key feature factor vector is a ix , the number of eigenvectors is m, the standard index B j The yth standard eigenfactor vector of jy , the number of eigenvectors is n;

[0202] According to the calculation results, retain the key indicators A i 5 standard indicators with the largest cosine values ​​B j ;

[0203] Fine matching: focus on indicator A i And the corresponding standard indicators Feature alignment obtains feature factor vectors with the same dimension and calculates the key index A i With standard indicators The weighted Mahalanobis distance is calculated based on the weighted Mahalanobis distance, and the key indicator A is matched according to the weighted Mahalanobis distance calculation result. i With standard indicators

[0204] Key Indicator A i The key characteristic factor vector is a i =(a i1 , a i2 , …, a ix , …, a in ), 1≤i≤m, m is the number of key indicators, n is the length of the feature vector, corresponding to the standard indicator The standard eigenfactor vector is j is an integer between 1 and 5;

[0205] The subjective weight is calculated by the hierarchical analysis method, and the expression is:

[0206]

[0207] where w z Standard eigenfactors calculated for the AHP The subjective weight of is the weight calculated by normalizing using the arithmetic mean method, is the weight calculated by normalization using the geometric mean method, and wz3 is the weight calculated using the eigenvalue method;

[0208] The objective weight is calculated by the CRITIC method, and the expression is:

[0209]

[0210] where w K Standard characteristic factors calculated by the CRITIC method The objective weight of j is the comparison strength, R j It is a conflict indicator;

[0211] The comprehensive weight is obtained based on the subjective weight and the objective weight, and the expression is:

[0212] w ij =τw z(ij)+vw K (ij)

[0213] where w ij is the standard characteristic factor The comprehensive weight of , τ and v are weight coefficients, τ+v=1;

[0214] Mean vector The covariance is Key Indicator A i With standard indicators The weighted Mahalanobis distance calculation is expressed as:

[0215]

[0216] Take the key index A corresponding to the minimum weighted Mahalanobis distance i and standard indicators is a pair of matching indicators to be compared, and m pairs of matching indicators to be compared are formed into an indicator group to be compared;

[0217] In the actual evaluation, taking the regional economic development level indicators as an example, the key indicators to be matched in the indicator feature library are: (Regional economic development level category: Indicator 1. Innovation and R&D momentum / good: R&D expenditure as a proportion of GDP -62, number of patent applications per million people -68, growth rate of the number of science and technology innovation enterprises -70; 2. Industrial optimization and upgrading / excellent: service industry value added as a proportion of GDP -85, high-tech industry output value as a proportion -88, industrial structure adjustment efficiency index -75).

[0218] In this embodiment, the method for constructing a regional standardized development data quantification model based on a neural network according to the indicator feature library includes:

[0219] The standard indicators, standard indicator categories, and feature representatives of the standard indicators in the indicator feature library are combined into a standard vector set, and the standard vector set is divided into a training set and a test set;

[0220] Input the training set into the XGBoost enhanced BP neural network for training;

[0221] Use the XGBoost model to predict the input training set, output new features and add them to the original training set, input the training set into the BP neural network for training, select the cross entropy loss function to evaluate the model, and select the SGD optimizer based on the evaluation to update the network parameters and output the development data quantization function;

[0222] The test set is used to evaluate the development data quantification function and build a regional standardized development data quantification model:

[0223]

[0224] Where S i is the quantitative model comparison score of the i-th indicator, S is the quantitative model comparison score of the user as a whole, and n i is the number of features of the ith indicator, m is the number of indicators of the user, is the category weight of the i-th indicator, is the indicator weight of the i-th indicator, I ij is the normalized value of the jth feature of the i-th indicator;

[0225] Input the indicator group to be compared into the regional standardized development data quantification model to output the comparison results between the indicator to be compared and the indicator feature library;

[0226] In actual evaluation, taking the regional economic development level indicators as an example, the category weights of regional economic development level Indicator weights of innovation input Weights of industrial structure indicators The comparison results of the key indicator groups are as follows:

[0227] Innovation input index: Innovation and R&D dynamics index = 38.115: 23.975

[0228] Industrial structure index: Industrial optimization and upgrading index = 31.325: 31

[0229] The comparison of regional standardized development data based on the quantitative model is completed by superimposing all categories of indicators.

[0230] Second, a regional standardized development data comparison system based on a quantitative model, including:

[0231] Data collection module: used to collect regional standardized development data, pre-process the development data, and transmit the development data to the screening module and the matching module; the development data includes historical indicators and indicators to be compared;

[0232] Screening module: used to screen the historical characteristic factors to obtain an indicator characteristic library;

[0233] Matching module: used for selecting the indicators to be compared according to the correlation to obtain the key indicators, and matching the indicators to be compared with the standard indicators of the indicator feature library to obtain the indicator group to be compared;

[0234] Machine learning module: used for inputting the standard feature library into the XGBoost enhanced BP neural network to obtain a regional standardized development data quantification model;

[0235] Comparison module: used to manage the development data, the indicator feature library and the indicator group to be compared, input the indicator group to be compared into the regional standardized development data quantification model, and output the comparison result between the indicator to be compared and the indicator feature library.

[0236] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for comparing regional standardized development data based on a quantitative model, characterized in that: The following steps are involved: S1. Obtaining standardized development data of a region and preprocessing the development data; the development data includes historical indicators and indicators to be compared; S2, extracting the features of the historical indicator and the indicator to be compared to obtain a historical feature factor and a comparison feature factor, performing feature screening on the historical feature factor to obtain a feature subset, and processing the feature subset to obtain an indicator feature library; the feature screening includes a first feature screening and a second feature screening; the feature subset includes a first feature subset and a second feature subset; S3, selecting the indicators to be compared according to the correlation to obtain key indicators, and matching the key indicators with the standard indicators in the indicator feature library to obtain an indicator group to be compared; S4. Construct a regional standardized development data quantification model based on a neural network according to the indicator feature library, input the indicator group to be compared into the regional standardized development data quantification model, and output the comparison result between the indicator to be compared and the indicator feature library.

2. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: The method for preprocessing the development data comprises: The development data is cleaned, deduplicated and normalized, the development data is classified using a decision tree algorithm, and the development data is labeled according to the classification results.

3. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: The method for screening the first feature comprises: Linear discriminant analysis is used to extract historical indicators and indicators to be compared to obtain historical characteristic factors and comparison characteristic factors; Define t ij as the j-dimensional feature corresponding to the i-th historical indicator, where 1 < i < m, 1 < j < n, m is the number of historical indicators, and n is the number of historical feature factors. The improved Harris hawk optimization algorithm is used to screen the features of the i-th historical indicator to obtain the first feature to form the first feature subset; The improved Harris Eagle optimization algorithm specifically includes: The historical characteristic factors are binary-coded, and the values ​​of the historical characteristic factors are 1 and 0 according to whether they are selected or not. The individual fitness of the eagle is calculated as follows: Where Fitness is the individual fitness of the eagle, (1-Accurcy) is the classification error rate, N is the feature length, ∝ is the accuracy of classification, and β=1-∝ is the weight of the selected feature in the fitness function; Introduce chaos mapping to update the rabbit escape energy formula: Where E k+1 is the updated escape energy of the rabbit, E0 is the initial energy of the rabbit’s escape, E k =2E0(1-t / T) is the rabbit's escape energy, t is the current iteration number, and T is the total iteration number; According to the updated bunny escape energy E k+1 And the Eagle's position update rule obtains the Eagle's position update X(t+1); Add normal distribution random numbers to perform Gaussian mutation on the eagle's position update: Where X(t+2) is the updated position of the eagle after Gaussian mutation, X best (t+2)=X(t+1)[1+N(μ,σ 2 )] is the value obtained by Gaussian mutation of the eagle's position X(t), N is a normally distributed random number, μ is the mean, σ 2 is the variance, f(·) is the fitness function, which indicates the fitness of the current position, rand is a random number, and p is the probability of survival of the fittest; KNN is used to establish a parcel feature selection model to evaluate the quality of feature subsets, where k=5. The individual fitness of the eagle is continued to be calculated and the position of the eagle group is updated. The first feature is output after the maximum number of iterations is reached, and the first features of m historical indicators are combined to obtain the first feature subset.

4. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: The method for screening the second feature comprises: Define the historical indicator set E = {e1, e2, ..., e m }, m is the number of historical indicators, the i-th historical indicator e i Corresponding to the historical characteristic factor set T i ={t i1 ,t i2 ,…,t in }, n is the number of historical characteristic factors; e i Corresponding sample, t ij Corresponding sample e i The j feature of the sample center circle with a radius of r is divided into a parts, and the multi-directional neighbor search method is used to find the k nearest hit points H of the same category. j and k different categories of recent miss points H j , the weights are calculated using the feature weight objective equation: Where W(t ij ) is the characteristic t ij weight, C is the category, Y is the category label set, Class(e i ) is the sample e i Get category, y(e i ) is the instance label, P(C) is the prior probability of category C, c is the number of categories, n is the number of features, diff(A,B,C) is the difference function, which indicates the difference between A in B and C, M l (Class(e i )) is the closest miss in category C, H l is the closest hit of the same type, u i is the smaller weight coefficient given to the sample when the similarity between the sample and the farthest miss is large, h i A larger weight coefficient is given to the sample when the similarity between the sample and the farthest hit is large; Filter feature selection by maximum Pierce correlation coefficient and maximum correlation distance: Where MPMD(t ij ) is t ij The feature weight of , t is the current iteration number, T is the total iteration number, a is a constant, PE(·) is the Pearson correlation coefficient, and the calculation rule is Calculate feature t ij Probability of being selected: Among them, PP(t ij ) is the characteristic t ij The probability of being selected, β is MPMD(t ij ) and W(t ij ), where β=1 is assumed to be equally important. The historical characteristic factors are selected according to the probability of being selected, and the second characteristics of m historical indicators are selected to obtain the second characteristic subset.

5. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: The method of processing the feature subset to obtain an indicator feature library includes: The intersection of the first feature subset and the second feature subset is taken as the feature representative, the unintersected parts S1 and S2 of the first feature subset and the second feature subset are taken to form the pending feature representative, the pending feature representative is screened twice to obtain the remaining feature representatives, and all feature representatives and corresponding historical indicators are combined to obtain the indicator feature library; Methods for secondary screening of candidate feature representatives include: Define the set \(T\) of undetermined characteristic representatives i =\(\{t i1 ,t i2 ,\cdots,t ij ,\cdots,t il \}\), where \(t ij \) is the \(j\)-dimensional characteristic corresponding to the \(i\)-th index, \(1 < i < m\), \(1 < j < l\), \(m\) is the number of indices, and \(l\) is the number of non-intersecting historical characteristic factors. Randomly assign an initial value to \(t ij \): Where rand∈(0,1) is a random value; The bird flock algorithm is used to screen historical characteristic factors for a second time and update the speed and position of the bird flock; The indicator is disturbed, and the expression is: Introducing Dunn's Index Calculate the fitness by combining the stability metric: Where Fitness is the fitness, 0.33 is the coefficient weight, errorrate is the classification error rate, dimension=l is the feature dimension, n is the original total number of features, minδ(S1,S2) is the minimum value of the historical feature factor distance between two disjoint sets, is the average value of the historical characteristic factor distance between two disjoint sets; According to the fitness calculation results, the undetermined feature representatives are screened to obtain the remaining feature representatives.

6. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: Methods for selecting indicators to be compared based on correlation to obtain key indicators include: Set the index i to be compared as the reference index X in turn i (j), i∈(0,m), m is the number of indicators to be compared, and the remaining indicators to be compared are set as the control indicators Y i (j), j∈(0,n) is the characteristic factor number of comparison index i, and n is the number of comparison characteristic factors; Calculate the relationship number between the reference indicator and the control indicator: where d i (k) is the reference index X i (j) and control index Y i (j) In the correlation coefficient of the comparison characteristic factor j, ρ is the resolution coefficient and is taken as 0.5; According to the weight w of the comparison feature factor j Calculate the correlation score between the reference indicator and the control indicator: Where R i X is the reference index i (j) and control index Y i (j) total relevance score, where a is the relevance score coefficient, and taking a = 10 changes the relevance score to a 10-point scale; The correlation of the indicators to be compared is evaluated by comprehensive grey correlation scores, and the indicators to be compared with high correlation are selected to obtain key indicators.

7. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: The method of matching the key indicator with the standard indicator of the indicator feature library to obtain the indicator group to be compared includes: Rough matching: defining key indicator A i =(a i1 ,a i2 ,…,a ix ,…,a im ) and standard indicator B in the indicator feature library j =(b j1 ,b j2 ,…,b jy ,…,b jn ), calculate key indicator A according to the indicator classification i All standard indicators of the same category B j The cosine value of is used to identify similar characteristic factor combinations according to the cosine value to obtain a rough matching index group; The key indicator A i All standard indicators of the same category B j The formula for calculating the cosine value of is: Where cos(A i ,B j ) is the key indicator A i With standard indicator B j Cosine value of eigenvector, key indicator A i The xth key feature factor vector is a ix , the number of eigenvectors is m, the standard index B j The yth standard eigenfactor vector of jy , the number of eigenvectors is n; According to the calculation results, retain the key indicators A i 5 standard indicators with the largest cosine values ​​B j ; Fine matching: focus on indicator A i And the corresponding standard indicators Feature alignment obtains feature factor vectors with the same dimension and calculates the key index A i With standard indicators The weighted Mahalanobis distance is calculated based on the weighted Mahalanobis distance, and the key indicator A is matched according to the weighted Mahalanobis distance calculation result. i With standard indicators Key Indicator A i The key characteristic factor vector is a i =(a i1 ,a i2 ,…,a ix ,…,a in ), 1≤i≤m, m is the number of key indicators, n is the length of the feature vector, corresponding to the standard indicator The standard eigenfactor vector is j is an integer between 1 and 5. The subjective weight is calculated by the hierarchical analysis method, and the expression is: where w z Standard eigenfactors calculated for the AHP The subjective weight of is the weight calculated by normalizing using the arithmetic mean method, is the weight calculated by geometric mean normalization, w z3 is the weight calculated by the eigenvalue method; The objective weight is calculated by the CRITIC method, and the expression is: where w K Standard characteristic factors calculated by the CRITIC method The objective weight of j is the comparison strength, R j It is a conflict indicator; The comprehensive weight is obtained based on the subjective weight and the objective weight, and the expression is: In ij =τw z (ij)+vw K (ij) where w ij is the standard characteristic factor The comprehensive weight of , τ and v are weight coefficients, τ+v=1; Mean vector The covariance is Key Indicator A i With standard indicators The weighted Mahalanobis distance calculation is expressed as: Take the key index A corresponding to the minimum weighted Mahalanobis distance i and standard indicator B j i is a pair of matching indicators to be compared, and m pairs of matching indicators to be compared are formed into an indicator group to be compared.

8. According to the method for comparing regional standardized development data based on quantitative model in claim 1, it is characterized in that: According to the indicator feature library, a regional standardized development data quantitative model based on a neural network is constructed, including: The standard indicators, standard indicator categories, and feature representatives of the standard indicators in the indicator feature library are combined into a standard vector set, and the standard vector set is divided into a training set and a test set; Input the training set into the XGBoost enhanced BP neural network for training: Use the XGBoost model to predict the input training set, output new features and add them to the original training set, input the training set into the BP neural network for training, select the cross entropy loss function to evaluate the model, and select the SGD optimizer based on the evaluation to update the network parameters and output the development data quantization function; The test set is used to evaluate the development data quantification function and to build a regional standardized development data quantification model. The indicator group to be compared is input into the regional standardized development data quantification model to output the comparison results between the indicators to be compared and the standard indicators in the indicator feature library.

9. A regional standardized development data comparison system based on a quantitative model, characterized in that: include: Data collection module: used for collecting regional standardized development data, preprocessing the development data, and transmitting the development data to the screening module and the matching module; The development data include historical indicators and indicators to be compared; Screening module: used to screen the historical characteristic factors to obtain an indicator characteristic library; Matching module: used for selecting the indicators to be compared according to the correlation to obtain the key indicators, and matching the key indicators with the standard indicators in the indicator feature library to obtain the indicator group to be compared; Machine learning module: used for inputting the standard feature library into the XGBoost enhanced BP neural network to obtain a regional standardized development data quantification model; Comparison module: used to manage the development data, the indicator feature library and the indicator group to be compared, input the indicator group to be compared into the regional standardized development data quantification model, and output the comparison result between the indicator to be compared and the indicator feature library.

Images