Performance evaluation of agricultural financial funds based on smart big data analysis

is paper attempts to build a performance evaluation system for agricultural nancial project expenditures to make the performance evaluation of nancial support for agriculture more maneuverable and provide a reference for the management of agricultural funds. Moreover, this paper combines the intelligent big data technology to construct the agricultural nancial fund performance evaluation system, and describes the characteristics of the relevance of geographic elements in the entire spatial region through global spatial autocorrelation. Simultaneously, this paper uses dierent spatial weight matrices to examine the rationality and robustness of the results. In addition, this paper introduces covariates and removes trends through regression methods. Further, taking into account the characteristics, scienticity and operability of agricultural nancial funds, the dimensions and structure of the performance evaluation of agricultural nancial expenditures are drawn up. Finally, this paper veries the eectiveness of this method through case studies, and gives several targeted suggestions.


Introduction
Agriculture is the foundation of the national economy, and nancial support for agriculture is an important direction of public nance expenditure. In recent years, all levels ofnance have increased their investment in agriculture-related funds year by year, and the scale of agricultural scal expenditure has risen sharply, and a stable growth mechanism for scal support of agriculture has been gradually established and improved [1]. In terms of the use of funds, the nancial support for agriculture has a direct bearing on the vital interests of the broad masses of farmers. erefore, it is a people's livelihood fund with a wide audience and a high degree of social attention, and it is of inestimable economic and political signi cance to manage and use scal funds to support agriculture [2]. However, judging from the current management and use of domestic nancial support funds for agriculture, it can be seen that there are many types, scattered use, random distribution, and frequent occurrences. ese problems are di cult to meet the needs of the development of the situation. erefore, it is imperative to carry out the performance evaluation of agricultural nancial expenditure, to re ect the input and output, cost and bene t of agricultural production to an operable and quanti able level, and to impel power into the cage of the system [3].
With the continuous growth of the population, the pressure on the demand for food has gradually emerged, and the living standards and quality of life of farmers have also been paid special attention. To this end, China has been committed to the development and progress of rural areas and agriculture. However, in the process of development, there are still many in uencing factors that restrict development, such as poor rural infrastructure, imperfect road tra c, imperfect water, electricity, and network communications, and a reduction in rural labor. As these in uencing factors have not been improved for a long time, it has led to the accumulation of more contradictions, the development of urban and rural areas has begun to become seriously unbalanced, the basic resources of rural society are scarce, and the income gap between urban and rural residents has become increasingly prominent. Facing the severe situation of rural and agricultural development, the Party Central Committee and the State Council have formulated a series of policies to strengthen the development of rural, agricultural, and peasants, and they have made rural, agricultural development and increasing peasants' income the top priority of government work. Among the influencing factors restricting the development of rural agriculture, weak rural basic production conditions, facilities, and limitations of rural farmers' development concepts have a great impact on rural development. In response to this situation, the state finance has invested a large amount of funds to support the development of agriculture and rural economy.
Fiscal expenditure is an important part of the national income distribution and redistribution. It is a resource allocation activity carried out by the state to meet the common needs of society, and it is an important aspect of public finance. With the continuous deepening of the reform of my country's fiscal budget management system and the establishment of the public finance framework system, strengthening the management and supervision of fiscal expenditures and improving the efficiency of the use of fiscal funds have become a very important task in fiscal management. At the same time, the establishment of a sound scientific fiscal expenditure performance evaluation system and the implementation of evaluation and control of fiscal expenditure performance are effective ways to improve the efficiency of fiscal expenditure. e method of national finance to support agriculture is generally direct investment, tax incentives and subsidies. e subsidies also include income subsidies, price subsidies, and financial discounts. Among them, the agricultural financial interest subsidy is an important part of the fiscal expenditure for supporting agriculture, and it is the national finance according to the agricultural development policy and industrial policy of the party and the government. It refers to the government granting a certain percentage of interest subsidies to certain specific loans within a certain period of time. It has an obvious financial leverage function, which can stimulate the investment enthusiasm of social funds and guide the investment direction of social funds. Under the background that all sectors of society pay more and more attention to the issue of fiscal expenditure efficiency, as a component of fiscal expenditure, the "tracking and testing" of agricultural fiscal discount interest and the evaluation of its use performance are of great significance to strengthening the management of public fiscal expenditure.

Related work
With the continuous progress of comprehensive agricultural development projects, in order to study the effects and efficiency of comprehensive agricultural development projects, many scholars try to evaluate the performance of agricultural projects by constructing a reasonable evaluation method and a scientific and reasonable indicator system. e unified indicator system framework cannot meet different evaluation needs, and many indicators do not truly reflect the performance of funds but are more directed toward work evaluation. erefore, the purpose of evaluation should become the starting point of the internal logic of evaluation [4]. Based on the theory of balanced scorecard and system theory, 12 specific and relatively independent indicators are selected from the four dimensions of beneficiaries, finance, management and operation and sustainability to form the project performance evaluation system [5]. e analytic hierarchy process is used to determine the weight of the performance evaluation index of land governance project funds [6]. Based on the process of agricultural comprehensive development project activities, the evaluation indicators are involved in two aspects: capital input and capital output [7].
From the perspective of project content classification, China's performance evaluation of agricultural development projects is done in accordance with the project. Due to the continuous refinement of land governance projects, the current research on land governance projects has gradually shifted from theoretical research on land governance project planning and feasibility analysis to empirical research. e method of combining theoretical analysis and expert consultation is used to establish an evaluation system to form a quantitative index evaluation and quantitative evaluation of the economic benefits after land consolidation [8]. e literature [9] established a provincial-based implementation index of domestic agricultural land quality, discussed domestic agricultural land changes, and re-division and evaluation of the domestic highyield, middle-yield, and low-yield fields and their regional distribution. e literature [10] analyzed and evaluated the status quo of low-yield farmland transformation in my country from the perspective of food security.
In the qualitative research on the management of comprehensive agricultural development funds, many domestic scholars analyze the management status and existing problems of comprehensive agricultural development funds from the aspect of fund management, and then put forward suggestions for optimizing management.
rough the research on the current situation of fund management for comprehensive agricultural development on the ground, it is found that there are problems in fund management such as low quality of financial personnel, irregular reimbursements, delayed fund audits, and lax project acceptance. e literature [11] proposed that the leadership should pay attention to the management of comprehensive agricultural development funds, strictly implement the county-level accounting system for fiscal funds, strengthen the audit of comprehensive agricultural development funds, strengthen the management of paid funds, and increase recovery efforts. rough a comprehensive analysis of the process of raising, allocating, appropriating, monitoring, and performance evaluation of agricultural comprehensive development funds, it was found that the contradiction between project funding needs and fund raising was prominent, and there were still system loopholes in the appropriation, use, and supervision of funds. Literature [12] proposes to improve the investment mechanism of comprehensive agricultural development, gradually expand the scale of investment, strengthen capital supervision, play a positive role in performance evaluation, and improve the quality of the comprehensive agricultural development fund management team. Taking the investment cycle data of a comprehensive agricultural development fund as an example, it analyzes from two aspects: the land governance capital chain and the industrialized management capital chain. Moreover, it reveals the problem of address changes and inaccurate budgets during the implementation of the agricultural development project due to inadequate preliminary work, which in turn led to poor circulation of agricultural development funds during the project implementation. e literature [13] proposed to speed up the construction of the financial system, implement supporting funds in place, strengthen rolling development, improve the quality of accounting personnel, and strengthen supervision, inspection and supervision. From the perspective of auditing, literature [14] used a combination of normative and empirical analysis to find many problems in the management of agricultural comprehensive development funds. ese problems are all caused by the serious principal-agent relationship in public management, and performance auditing is an effective measure to solve this problem. e management of comprehensive agricultural development funds is out of touch with project management. Since project management and fund management are not the same department, each department performs its own duties, and there is no complete right to know each other, resulting in poor fund allocation. e literature [15] suggested that the relevant departments of comprehensive agricultural development should correspondingly strengthen project plan management, make more detailed budgets for projects, improve project and fund declaration procedures, and strengthen communication and coordination and information sharing among relevant departments.

Smart big data inspection algorithm
Global spatial autocorrelation describes the description of the characteristics of the relevance of geographic elements in the entire spatial region. Global Moran's I is a widely used global spatial autocorrelation statistic. e calculation formula is as follows [16]: (1) Among them, S 0 � n i�1 n j�1 w ij , x � n i�1 (z i /n), and n is the total number of observation variables, (z i − x) is the deviation of the observation value on the i-th spatial unit from the average value, and w ij is the weight of the spatial observation variables i and j. i and j represent two different regions. e spatial weight matrix w is defined as: Among them, the spatial weight matrix w is a real symmetric matrix. e elements on the main diagonal indicate the distance of the same area, and W 11 � . . . . � W nn � o indicates that the distance between the same area is 0. e most commonly used distance function is "adjacent", that is, if the area i and the area j have a common boundary, when it is expressed by the indicator function (also called the Heavyside function), w ij � l, otherwise, w ij � 0. For the convenience of description, the vector Z can be used to represent the attribute variable of the space object, When applying the model, the weight matrix should be subjected to "row standardization" processing. at is, each element in the matrix (denoted as w ij ) is divided by the sum of the elements in its row to ensure that the sum of the elements in each row is 1. e advantage of implementing row standardization is that if the row standardization matrix W is multiplied by z, the average value of the attribute values around each area can be obtained. e calculation formula for row standardization is as follows [17]: Similarly, for the space-time weight matrix, W st X is called the "spatial lag" of X, that is, the average value of X's surrounding attribute values. Since the sum of the elements in each row is 1, this means that the sum of the influences of its neighbors on the area i must be equal to the sum of the influences of its neighbors on the area j (any i ≠ j).
ere are two issues worthy of attention when carrying out the Moran's I spatial autocorrelation test in the traditional space. First, the value of Moran's I has a strong correlation with the spatial weight matrix W. If the weight matrix is not properly selected or set, unreasonable or false spatial autocorrelation results will be obtained. Second, the core component of Moran's I model is (z i − x)(z j − x), which implicitly assumes that the expected value of z i n i�1 is constant. erefore, the premise of using this model is that there is no trend in the attribute value of the geospatial object, that is, it is stable in the spatial dimension.
In order to make the calculation results of the model more in line with reality, the appropriate spatial weight matrix should be selected first, or different spatial weight matrices should be used to examine the rationality and robustness of the results. Secondly, covariates can be introduced, the trend can be removed by the regression method, and Moran's I statistical test can be performed on the residual items. When performing statistical inference, it is necessary to make assumptions about the distribution of the spatial variable Z in advance. ere are generally two assumptions: the first is to assume that the spatial variable Z obeys a normal distribution, and the second is to use randomization to get the approximate distribution of Z when the distribution is unknown. Under the two assumptions of normal distribution or random uniform distribution, the expected value and variance of the spatial autocorrelation coefficient are obtained and tested [18].
Under the assumption of normal distribution, the expected value and variance of Moran'sI are:

Mobile Information Systems
Under the assumption of random uniform distribution, the expected value and variance of Moran's I are: Here When n is large enough, Moran's I expectation is close to zero. e value range of Moran's I is [−1, 1]. e closer the value is to ±1, the higher the spatial autocorrelation; when the value of Mroan's I is greater than its expected value, the adjacent area or position. e attribute values of both tend to be similar, showing a positive autocorrelation: when the value of Moran's I is less than its expected value, the variable attribute values in adjacent areas or locations tend to be different, showing a negative autocorrelation. When Moran's I>0, it indicates a positive correlation, that is, areas with high attribute values are clustered with areas with high attribute values (high-high aggregation, or IHIH aggregation for short), and areas with low attribute values are clustered with areas with low attribute values. (Low low poly, LL aggregation for short), as shown in Figure 1(a). e closer Moran's I is to 1, the higher the degree of spatial aggregation and the smaller the overall spatial difference; on the contrary, when Moran's I <0, it indicates a negative correlation, and the attribute values of the observation unit present a discrete spatial pattern, as shown in Figure 1(b) Shown. e closer Moran's I is to −1, the greater the overall spatial difference. When Moran's-0, the observed variables are independent of each other, and there is no spatial autocorrelation in the attribute values of the observation units, and they are randomly distributed in space. At this time, Moran's I is close to −1/(n−1), as shown in Figure 1(c) as shown [19].
If it is assumed that the observed variable does not have spatial autocorrelation in space, but presents a random distribution, in the statistical test, the standardized Moran's I value is generally used, namely [20]: Among them, E(I) is Moran's I expected value of spatial autocorrelation, var(I) and S(I) represent the variance and standard deviation of spatial autocorrelation Moran's I, respectively. Z(I) is a multiple of the standard deviation. It is used to test the size of the spatial autocorrelation.
At the 0.05 confidence level, |Z(I)| � 1.96. Z(I) > 1.96 means: there is a significant positive correlation between the observed variables.
at is, high observations and high observations have spatial aggregation (HH), and low observations and low observations have spatial aggregation (LL), which presents a spatial aggregation distribution. |Z(I)| < −196 means: there is a significant negative correlation between observations, that is, high observations and low observations are clustered together (HL), and low observations and high observations are clustered together (LH), which presents an abnormal spatial distribution. |Z(I)| < 1.96 means: the correlation of the spatial observation variable is not significant, and it presents an independent random distribution in the region. e global spatial autocorrelation statistics are based on the implicit assumption of spatial stationarity, that is, the expected value and variance of the attribute values of the observed variables at all positions are constant. However, most of the geospatial processes are non-stationary, especially when the amount of data is very large, the requirement of spatial stability becomes very unrealistic. From the perspective of the research area, there are often different degrees of spatial autocorrelation among spatial elements. is phenomenon is also called Spatial heteroge-neity. Local spatial autocorrelation statistics can identify different spatial association patterns (or spatial aggregation patterns) that may exist at different spatial locations. erefore, observing the local instability at different spatial locations can find spatial heterogeneity between data. It is this spatial heterogeneity that provides the basis for the classification or division of spatial patterns.
In essence, the local space Moran's I model decomposes the global space Moran's I model into various regional units. For a certain space i, local Moran's I is expressed as follows [21]: When I t > 0, it means that the spatial objects with similar observation values are in a clustered state in space, which can be high-value clustering or low-value clustering, and this phenomenon is called "cluster". Conversely, when I0, it means that the spatial objects with dissimilar observations gather in space, and this phenomenon is called "outlier". e so-called "cluster" or "outlier" only represents the deviation state of the attribute value of the spatial unit, and does not represent the geographical distribution of the spatial objects. e exact distribution form of local Moran's I statistic is generally unknown, and its test usually adopts conditional randomization or random arrangement method. Conditional randomization refers to the empirical distribution function that fixes the observation value of the position and arranges other observation values randomly in the entire space position.Under the assumption of complete randomness, the mean and variance of I t are respectively; In the above formulas (8) and (9), e constructed test statistics are: In addition, Moran scatter plots can also identify spatial anomalies and local instabilities. e Moran scatter diagram describes the correlation between the observation vector Z of a certain unit and its spatial lag vector WZ (that is, the weighted average of the observation values of the surrounding units of the space unit). For example, the Moran scatter corresponding to (Z, WZ) In the dot plot, the slope corresponding to the straight line fitted by the scatter points is the size of the global Moran'sI value. Because the data set (Z, WZ) is processed by row standardization, the results at different times are comparable Sex.
e Moran scatter diagram has 4 quadrants, which correspond to the 4 local spatial connection forms between the regional unit and its neighbors: the first quadrant represents the high observation value of the regional unit and the surrounding area is high (HH); the second quadrant represents e area around the area unit with low observation value is high value (LH): the third quadrant represents the area around the area unit with low observation value is the same as the low value (LL); the fourth quadrant Mobile Information Systems represents the area around the area unit with high observation value It is a low value (HL). Compared with the local Moran's I, the advantage of the Moran scatter plot is that it can further specifically distinguish which spatial association mode among the HH, LL, HL, and LH the observation values between the regional unit and its adjacent units belong. erefore, according to the different quadrants of the data points (Z, WZ) in the Moran scatter diagram, it is possible to identify what kind of spatial distribution the observed value of the research object of the regional unit belongs to.

Case analysis of performance evaluation of agricultural financial funds
is article takes the A region as an example to conduct a case study. Taking into account the characteristics of agricultural financial funds, scientificity and operability, this paper draws up the dimensions and structure of agricultural financial expenditure performance evaluation, as shown in Figure 2, which includes three dimensions: project plan, project management, and project performance.
According to the statistical data of expert judgment, we first construct the judgment matrix A of the first-level index, as shown in Table 1, and carry out the consistency test.
We get that W(A1) �0.1841, W(A2) �0.3062, W(A3) �0.5113, CI�0.011, CR�0.0306. erefore, CR<0.1, which proves that the judgment matrix has good consistency, can pass the test, and the weight distribution of each first-level indicator is reasonable, as shown in Table 2. e first step is to construct a paired comparison matrix B by judging the importance of the two factors. Numbers are used as a scale to judge the relative importance of two factors and the two factors are compared one by one to form a paired comparison matrix, as shown in Table 3. e second step is to calculate the weights through the consistency check method and judge the rationality of the index weight settings. Using YAAHP software, the following results are obtained: W(B1) �0.1634, W(B2) �0.2970, W(B3) �0.5396, CI�0.0046, CR�0.0079. We define the consistency ratio as CR�CI/RI. When analyzing the paired comparison matrix B, we take 0.1 as the boundary value. When its value is below the boundary value, it is considered to have a certain degree of consistency. Otherwise, it is necessary to reconstruct the comparison matrix B to adjust the relative importance of each index. CR < 0.1, so the hierarchical ranking passes the consistency test. e consistency test and index score calculation table are shown in Table 4 below.
We choose Area A from January 2020 to May 2021 for 26 projects, involving agricultural technology promotion, green food, agricultural information, planting, breeding, agricultural talent training, and agricultural land planning. Due to the large number of projects, in order to facilitate the visual display of related data and layout, 26 projects are replaced by projects 1 to 26.  From the perspectives of self-evaluation and third-party evaluation of the project unit, performance evaluation empirical studies are carried out, and there are still some differences in the evaluation scores. Unit self-assessment is to better obtain basic information, self-assessment reports, key supporting materials, etc. of various fund authorities and user units for reference. By carrying out third-party evaluation and combining with the self-evaluation results of the project unit to form a comparison, it will help to further discover problems and differences and make up for the lack of self-evaluation, as shown in Tables 5 and 6, and Figures 3 and 4. From the above experimental research, it can be seen that when each unit conducts project self-evaluation, there is a tendency to score in a good direction as much as possible. However, the performance evaluation carried out from the      Table 6: Comparison of self-assessment and empirical performance levels. perspective of a third party can more objectively and reasonably examine the actual overall performance of the agricultural financial fund project.

Conclusion
e performance of agricultural financial funds has a bearing on agriculture, rural areas and farmers, as well as the implementation of rural revitalization strategies. By strengthening the performance evaluation of agricultural financial funds, the efficiency of the use of funds can be improved, and performance management can better serve budget management. It is necessary to promote budget units to establish budget performance management concepts, effectively improve the important role of performance evaluation in daily management, ensure the safe and efficient use of financial funds, and continuously expand performance evaluation models. At the same time, it is necessary to improve the quality of performance evaluation, make full use of the results of performance evaluation, and continuously improve the efficiency of the use of financial funds, so as to achieve the goals of promoting the change of the government's governance concept, strictly controlling the budget, establishing a responsible government, and providing effective public services. is paper combines smart big data technology to evaluate the performance of agricultural financial funds. Moreover, this paper proposes several improvement measures including improving the performance evaluation organization system, attaching importance to the preliminary project demonstration and indicator design, strengthening the process supervision of project performance and funds, and expanding the output and benefits of agricultural financial funds.

Data Availability
e experimental data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e author declared that there are no conflicts of interest regarding this work.