Comprehensive analysis of risk factors in Internet agricultural finance based on neural network model

Abstract

Agricultural industrialization is a major reform and practice in the process of agricultural development and requires theoretical guidance. However, the current theoretical research on financial support for the development of agricultural industrialization is insufficient, which to a certain extent seriously affects the development speed of agricultural industrialization. This paper studies the nature of the part and tail probability of dependent random variable sequences with different distributions, and focuses on the random variable sequences with wide dependent structures, and obtains the relevant probability estimation formulas. At the same time, this paper also considers the application of the main results in complete convergence. Moreover, based on the research on the nature of dependent random variable sequences, the dependent risk model is discussed, which combines Internet finance with the development of agricultural industrialization. In addition, this article uses agricultural industrialization theory and Internet finance theory to study the support of Internet finance for the development of agricultural industrialization in my country. The research results show that the model constructed in this paper has a certain effect.

Keywords

Neural network internet finance agriculture risk factors

1 Introduction

From the current domestic research, it can be seen that the risk control work of my country’s financial institutions mainly starts with traditional banks. At the beginning, the risk control was mainly carried out by manual means of experience control, and the bank formulated corresponding rules to rate personnel to control the risk of loan default. China Construction Bank conducts personnel ratings based on the age, education, income, occupation and other indicators of credit card users, and grants cardholders corresponding overdraft lines based on different ratings to control risk [1]. In the Internet finance industry, the risk control of lending mainly depends on the accurate assessment of the credit of loan users. The experience and methods of traditional financial industry, especially traditional commercial banks, in solving this problem are worth learning from the entire Internet financial industry. The traditional financial industry, especially some commercial banks, has experienced the impact of many economic crises, and its internal financial institutions have relatively mature theories and system structures for user credit review. In the bank’s credit evaluation methods for user loans, a more effective method is to build a score card model. There are mainly two standard scorecard models, one is the application scorecard model, and the other is the behavioral scorecard model. The application scorecard model refers to the judgment of the default risk of the user’s loan application to screen the user. The behavioral score card model refers to the management of the entire loan cycle for users who have passed the loan application. Among the scorecard models, the FICO scorecard model in the United States is relatively mature and widely used. This model is proposed by the American Personal Consumer Credit Rating Enterprise on the basis of millions of data. Based on the user’s credit, morality, and ability to pay, the model refines each dimension into multiple grades, calculates the score of each grade, and finally sums the weighted scores of each dimension to obtain the user’s total score. Moreover, the model judges the user’s credit through corresponding scores, thereby providing financial institutions with a basis for reviewing user loan applications and assisting them in making decisions. The higher the score, the better the user’s credit. With the development of Internet technology, the application of Internet technology in the financial field has prompted the rise of Internet finance, among which risk control is widely used in Internet financial lending [2]. Compared with traditional bank loans, this loan model has obvious advantages. First, it solves the problems caused by the asymmetry of information between borrowers and lenders very well. Second, this loan method is not only convenient and fast, but also provides a channel for capital providers to increase capital and solves the problem of financing difficulties for those who need capital. Of course, while it is convenient and efficient, it is also accompanied by the risks of overdue loans and personal information leakage. Using Internet technology to control risks is a core issue that needs to be resolved in the Internet finance field in the future. Compared with bank lending, Internet finance lending faces more complex people and faces a higher default rate. By referring to the risk control methods and systems of the traditional financial industry, algorithms such as machine learning and deep learning are used for big data analysis and mining, and there are already many mature experience and application systems in loan risk control and default identification [3].

2 Related work

Regarding the role of the score card model, the literature [4] pointed out that the FICO score card model has strong flexibility and applicability. It can build a model for risk control based on user data. As user data changes, FICO’s scoring results will also change accordingly. Therefore, the FICO scorecard model can accurately and continuously measure the user’s credit status, which can help financial institutions to carry out risk control well. The literature [5] used scorecard model, logistic regression algorithm and decision tree model to model the user consumption data of entertainment clubs to evaluate their credit and determine whether they will be overdue after borrowing. By comparing the prediction results of the scorecard model, the logistic regression algorithm and the decision tree model, it is found that the readiness of the three machine learning models is similar, but the scorecard model is more convenient and simple in operation.

In order to solve related problems more effectively, the literature [6] used logistic regression model, support vector machine and ensemble learning model random forest to train more than 300,000 pieces of data, and realized an automated system that can process and model data. The literature [7] predicted Internet financial data by constructing a regression model based on the kernel function, and found that this model is superior to existing methods in terms of risk control. By studying the difference between the risk control of loan default in the Internet financial industry and the risk control of traditional financial industry in loan default, the literature [8] found that the Internet financial industry has borrowed many methods from the traditional financial industry in terms of risk control. The literature [9] used the integrated learning model random forest to predict the loan default of online loan users, so that the probability of user default is quantified, and a better credit rating system is established. The literature [10] studied the experience of traditional small loans in risk control and the role of the bank’s credit evaluation system in risk control. On this basis, it re-increased the risk factors in online lending and built a new risk control model using BP neural network. The literature [11] took a certain car loan platform as the research object, and constructed a logistic regression model and a Cox model to study the factors that affect users’ default after borrowing. It was found that the factors that are positively correlated with the loan default rate are: foreign household registration and historical default records, etc., and the factors that are negatively related to the loan default rate are: relatives’ knowledge, high credit ratings and other factors. The literature [12] built a loan default early warning model based on multi-dimensional features, so that the network lending platform reduces the frequency of loan defaults in terms of bad debts and fraud. By studying the development of Internet finance, the literature [13] proposed that the collection of credit information is conducive to the control of the default risk of Internet finance loans. The literature [14] conducted empirical analysis by constructing a structural equation model to study which factors of users affect their borrowing on Internet financial platforms, and found that users’ social capital, economic status, and credit rating have significant effects. By studying the risk control system of Ant Financial under Alibaba, the literature [15] found that the platform builds a deep learning model to train users’ multi-dimensional feature data and multi-level indicator evaluation data to control the risk of loan default. By studying the loan risk of commercial banks, the literature [16] pointed out that on the basis of the evaluation of the borrowing user’s level, the profitability, operation ability and solvency of the user can be used as features that can reflect the solvency of the enterprise, and a logistic regression model can be constructed to predict the risk of loan default. After empirical research, the literature [17] proposed to use factor analysis to select features that have strong ability to select dependent variables for modeling, which can effectively solve the problem of model over-fitting due to too many features when building the model.

According to the transaction cost theory, the literature [18] measured the development level of my country’s Internet finance from three levels of security, liquidity and profitability. Based on the theory of financial functions, the literature [19] constructed a set of indicator systems to measure the efficiency and equilibrium level of Internet finance development from four levels: payment methods, resource allocation, risk control and information processing. Moreover, it uses methods that conform to the system order and synergy model to evaluate the overall development level of China’s Internet finance from 2007 to 2014. The literature [20] constructed a set of evaluation index system using 8 provinces and cities with relatively developed Internet finance in my country as research samples, and used quantitative analysis methods to evaluate the development level of Internet finance in my country in 2014. The literature [21] used cluster analysis and factor analysis to make an empirical analysis of the spatial differences and development levels of my country’s Internet finance. The results show that my country’s Internet finance is still in its initial stage of development, and there are obvious regional differences. The degree of regulation of Internet finance development depends on the optimization of Internet finance development measures. The literature [22] used a logistic regression model to calculate the loan default probability of people with different credits, which provides ideas for the further improvement of the credit scoring card model.

3 Compound dependent risk model and exquisite large deviation of total claims

In the compound dependent risk model, the claim amount for the j (j > 1)-th claim caused by the i (i ⩾ 1)-th accident is recorded as X_ij. At the same time, {X_ij, j> 1, i ⩾ 1 } is a sequence of non-negative random variables with a finite mean. The time interval of accidents constitutes a series of random variables {θ_i, i⩾ 1 }, and {θ_i, i⩾ 1 } is a series of independent and identically distributed random variables with positive values and a finite mean. Then, from this, an update counting process is constituted: $N (t) = sup {n ⩾ 1 : τ_{n} = \sum_{i = 1}^{n} θ_{i} ⩽ t}, t ⩾ 0$ (1)

Its mean function is λ (t) = EN (t) , t ⩾ 0, and when t→ ∞ is satisfied, there is λ (t)/λt → 1. In this model, the number of claims caused by successive accidents is recorded as {Y_i, i⩾ 1 }, {Y_i, i⩾ 1 } is a sequence of independent identically distributed random variables with non-negative integer values, and has a finite mean v. We assume that the random variable sequence {Y_i, i⩾ 1 } is bounded, that is, there is a finite integer h > 0 such that Y_i ⩽ h, i ⩾ 1. Therefore, the total claims until time t ⩾ 0 is expressed as: $S (t) = \sum_{i = 1}^{N (t)} \sum_{j = 1}^{Y_{i}} X_{ij}$ (2)

For the standard updated risk model, the early research on the exquisite large deviation of total claims and the probability of bankruptcy mainly considers the situation that the claim amount and the time interval of the claim are all independent and identically distributed random variables. In fact, the independence assumption is no longer applicable to insurance practice. Therefore, more and more scholars have begun to study risk models with dependent assumptions[23].

This section studies the exquisite large deviation of the total claims in the compound risk model, that is, the exquisite large deviation of random and S (t). At the same time, it prepares for the asymptotic analysis of the probability of ruin in the next section. For the above risk model, an important assumption of the above results is that the amount of claims, the time interval between accidents and the number of claims generated by each accident are mutually independent. However, in actual problems, if an excessively large accident occurs for a period of time and generates more claims, people will raise their awareness of safe production, and the time interval for another accident will become longer. This shows that there will be a certain dependency between the number of claims caused by accidents and the time interval between accidents.

We assume that { (θ_i, Y_i) , i⩾ 1 } is a sequence of pairs of independent and identically distributed random variables. Moreover, we set θ and Y to be the general random variables of the random variable sequence {θ_i, i⩾ 1 } and {Y_i, i⩾ 1 } respectively, and satisfy that for all t ∈ [0, ∞) and any 1 ⩽ k ⩽ h, there are: $P (θ > t | Y = k) ⩽ p (θ^{*} > t)$ (3)

Among them, θ^∗ is a non-negative random variable, and it is independent of any other random variable. We set $θ_{1}^{*}$ to be a positive random variable, and it is independent of any other random variable, so that for all t ∈ [0, ∞) and any 1 ⩽ k ⩽ h, there is [24]: $P (θ_{1}^{*} > t) = P (θ_{1} > t | Y_{1} = k)$ (4)

Moreover, we set: $\begin{matrix} τ_{1}^{*} = θ_{1}^{*}, \\ τ_{n}^{*} = θ_{1}^{*} + \sum_{i = 2}^{n} θ_{1} \\ n ⩾ 2 \end{matrix}$ (5)

In this way, it constitutes a delayed update counting process: $N^{*} (t) = sup {n ⩾ 1 : τ_{n}^{*} ⩽ t}, t ⩾ 0$ (6)

Therefore, for all integers n ⩾ 1 and 1 ⩽ k ⩽ h, there is the following formula: $\begin{matrix} P (N^{*} (t) = n) = P (τ_{n}^{*} ⩽ t, τ_{n + 1}^{*} > t) \\ = \int_{0}^{\infty} P (\sum_{i = 2}^{n} θ_{i} ⩽ t - u, \sum_{i = 2}^{n + 1} θ_{i} > t - u) P (θ_{1}^{*} \in du) \\ = \int_{0}^{\infty} P (\sum_{i = 2}^{n} θ_{i} ⩽ t - u, \sum_{i = 2}^{n + 1} θ_{i} > t - u) P (θ_{1}^{*} \in du | Y_{1} = k) \\ = P (\sum_{i = 1}^{n} θ_{i} ⩽ t, \sum_{i = 2}^{n + 1} θ_{i} > t | Y_{1} = k) \\ = P (N (t) = n | Y_{1} = k) \end{matrix}$ (7)

We consider the situation where the claim amount is positive and equally distributed. At the same time, we set that the time interval between accidents and the number of claims each constituted by the random variable series are independent and identically distributed. Moreover, for each i ⩾ 1, there is a dependency structure between the time interval θ_i between the occurrence of the i-th accident and the number of claims Y_i caused. When the claim amount {X_ij, j> 1, i ⩾ 1 } satisfies the dependency structure and the common distribution F satisfies F ∈ L ∩ D, the asymptotic lower bound of the refined large deviation of the total claim S (t) is obtained. When the claim amount {X_ij, j> 1, i ⩾ 1 } satisfies the wide upper quadrant dependent (WUOD) structure and there is β > 1 such that ${EX}_{1}^{β} < \infty$ , and the common distribution F satisfies F ∈ C, the asymptotic upper bound of the refined large deviation of $\sum_{i = 1}^{N (t)} \sum_{j = 1}^{Y_{i}} (X_{i j} - E X_{1})$ is obtained.

The situation where there is a dependency between the time interval of accident occurrence and the number of claims is continued to be considered, that is, for each i ⩾ 1, there is a dependency structure between the time interval θ_i of the i-th accident occurrence and the number of claims Y_i caused. For each i ⩾ 1, {X_ij, j> 1 } is a sequence of random variables with the same distribution F_i ∈ D, and the random variable sequence {X_ij, j> 1, i ⩾ 1 } composed of the claim amount satisfies a certain dependent structure. Under the above conditions, we study the exquisite large deviation of the total claim S (t). To this end, we first give the following assumptions:

Hypothesis 1: The claim amount {X_ij, j> 1, i ⩾ 1 } is a sequence of non-negative random variables, and for α > 0, there is [25]: $lim_{n \to \infty} sup_{x ⩾ α n} sup_{1 ⩽ i ⩽ n} sup_{1 ⩽ j ⩽ l ⩽ h} xP (X_{ij} > x | X_{il} > x) = 0$ (8)

for $lim_{n \to \infty} sup_{x ⩾ α n} sup_{1 ⩽ i \neq k ⩽ n} sup_{1 ⩽ jh, 1 ⩽ l ⩽ h} xP (X_{ij} > x | X_{kl} > x) = 0$ (9)

And, for each i ⩾ 1, the random variable sequence {X_ij, j> 1 } has the same distribution F_i.

Hypothesis 2: Time interval between accidents, {θ_i, i⩾ 1 } is a sequence of positive random variables with a finite mean λ^-1. The number of claims {Y_i, i⩾ 1 } is a sequence of non-negative integer random variables with a finite mean v. { (θ_i, Y_i) , i⩾ 1 } is a sequence of pairs of independent and identically distributed random variables, and (θ, Y) satisfies the dependency structure. In addition, {X_ij, j> 1, i ⩾ 1 } is independent of {θ_i, i⩾ 1 } and {Y_i, i⩾ 1 }, respectively.

First, we study the asymptotic lower bound of the refined large deviation of the total claim. In this case, for distributions F_i, i ⩾ 1, we need to use the following assumptions:

Hypothesis 3: There is a distribution function F such that F_i, i ⩾ 1 satisfy the following relationship: $0 < M_{l} : = \underset{x \to \infty}{lim inf} inf_{i ⩾ 1} \frac{{\bar{F}}_{i} (x)}{\bar{F} (x)} ⩽ \underset{x \to \infty}{\lim sup} sup_{i ⩾ 1} \frac{{\bar{F}}_{i} (x)}{\bar{F} (x)} = : M_{U} < \infty$ (10)

We can get that if F_i ∈ D, i ⩾ 1, then there is F ∈ D, and for any y > 1 and i ⩾ 1, there is: $\frac{M_{L}}{M_{U}} {\bar{F}}_{i *} (y) ⩽ {\bar{F}}_{*} (y) ⩽ \frac{M_{U}}{M_{L}} {\bar{F}}_{i *} (y)$ (11)

Therefore, for any i ⩾ 1, there are $J_{F}^{+} = J_{F_{i}}^{+}$ and $\frac{M_{L}}{M_{U}} L_{F_{i}} ⩽ L_{F} ⩽ \frac{M_{U}}{M_{L}} L_{F_{i}}$ . Under the conditions of hypothesis 1, hypothesis 2, and hypothesis 3, we obtain the asymptotic lower bound of the refined large deviation of the total claim.

Theorem 1: When considering the total claim, if hypothesis 1, hypothesis 2 and hypothesis 3 are all established, and F_i ∈ D, i ⩾ 1, for any γ > 0, there is: $P (S (t) > x) > v \sum_{i = 1}^{λ_{t}} L_{F_{i}} {\bar{F}}_{i} (x)$ (12)

It is consistent for all x ⩾ γt. Equivalently, there is: $\underset{t \to \infty}{lim inf} inf_{x ⩾ γ t} \frac{P (S (t) > x)}{v \sum_{i = 1}^{λ_{t}} L_{F_{i}} {\bar{F}}_{i} (x)} ⩾ 1$ (13)

When {X_ij, j> 1, i ⩾ 1 } is a sequence of identically distributed random variables, we can directly obtain the following deduction from theorem 1.

Corollary 1: When considering the total claims, if assumption 2 is true, and the claim amount {X_ij, j> 1, i ⩾ 1 } is a sequence of non-negative random variables with the same distribution F. At the same time, it satisfies the hypothesis 1, and F ∈ D, for any γ > 0, there is: $P (S (t) > x) > v λ_{t} L_{F} \bar{F} (x)$ (14)

It is consistent for all x ⩾ γt. Equivalently, there is: $\underset{t \to \infty}{lim inf} inf_{x ⩾ γ t} \frac{P (S (t) > x)}{v λ_{t} L_{F} \bar{F} (x)} ⩾ 1$ (15)

The asymptotic upper bound of the exquisite large deviation of the total claim. In this case, we consider that the claim amount {X_ij, j> 1, i ⩾ 1 } has a pairwise negative quadrant dependent structure.

Hypothesis 4: The claim amount {X_ij, j> 1, i ⩾ 1 } is a sequence of non-negative random variables and satisfies the pairwise negative quadrant dependency structure, that is, for all x ⩾ 0 and y ⩾ 0, there is

$\begin{matrix} P (X_{ij} > x, X_{ij} > y) ⩽ P (X_{ij} > x) \\ P (X_{ij} > y), i, j, l ⩾ 1, j < l \end{matrix}$ (16)

and

$\begin{matrix} P (X_{ij} > x, X_{kl} > y) ⩽ P (X_{ij} > x) \\ P (X_{kl} > y), i, j, k, l ⩾ 1, i \neq k \end{matrix}$ (17)

At the same time, for each i ⩾ 1, the random variable sequence {X_ij, j> 1 } has the same distribution F_i.

Hypothesis 5: For all i ⩾ 1, F_i ∈ D. Further, we assume that for any ɛ > 0, there are two constants ω₁ = ω₁ (ɛ)> max { h, 1 } and x₁ = x₁ (ɛ) > 0 that are independent of i, so that for all i ⩾ 1, 1 ⩽ ω ⩽ ω₁ and x ⩾ x₁. $\frac{{\bar{F}}_{i} (ω x)}{{\bar{F}}_{i} (x)} ⩾ L_{F_{i}} - ɛ$ (18)

Or equivalently, for any ɛ > 0, there are two constants 0< ω₂ = ω₂ (ɛ) < min { h^-1, 1 } and x₂ = x₂ (ɛ) > 0 that have nothing to do with i, so that for all i ⩾ 1, ω₂ ⩽ ω ⩽ 1 and x ⩾ x₂. $\frac{{\bar{F}}_{i} (ω x)}{{\bar{F}}_{i} (x)} ⩽ L_{F_{i}}^{- 1} + ɛ$ (19)

Under the conditions of hypothesis 2, hypothesis 3, hypothesis 4 and hypothesis 5, we obtain the asymptotic upper bound of the refined large deviation of the total claim.

Theorem 2: When considering total claims, if hypothesis 2, hypothesis 3, hypothesis 4 and hypothesis 5 are all true, and there is some $p > J_{F}^{+}$ that makes ${EX}_{ij}^{2 p + 1} < \infty, i, j ⩾ 1$ , there is γ₁ > 0 that makes for any γ ⩾ γ₁. $P (S (t) > x) < v \sum_{i = 1}^{λ_{t}} L_{F_{i}}^{- 1} {\bar{F}}_{i} (x)$ (20)

It is consistent for all x ⩾ γt. Equivalently, there is: $\underset{t \to \infty}{lim sup} sup_{x ⩾ γ t} \frac{P (S (t) > x)}{v \sum_{i = 1}^{λ_{t}} L_{F_{i}}^{- 1} {\bar{F}}_{i} (x)} ⩽ 1$ (21)

If {X_ij, j> 1, i ⩾ 1 } is a sequence of identically distributed random variables and F ∈ D, then hypothesis 3 and hypothesis 5 are both true. Therefore, we can get the following inference from theorem 2.

Corollary 2: When considering the total claims, if hypothesis 2 holds, the claim amount {X_ij, j> 1, i ⩾ 1 } is a sequence of non-negative random variables with the same distribution F, and it satisfies the hypothesis 4, and F ∈ D and there is a certain $p > J_{F}^{+}$ such that ${EX}_{ij}^{2 p + 1} < \infty, i, j ⩾ 1$ , then there is γ₁ > 0 such that for any γ ⩾ γ₁, there is: $P (S (t) > x) < v λ_{t} L_{F}^{- 1} \bar{F} (x)$ (22)

It is consistent for all x ⩾ γt. Equivalently, there is: $\underset{t \to \infty}{lim sup} sup_{x ⩾ γ t} \frac{P (S (t) > x)}{v λ_{t} L_{F}^{- 1} \bar{F} (x)} ⩽ 1$ (23)

When F ∈ C, we have L_F = 1. From corollary 1 and corollary 2, the following corollaries can be obtained.

Corollary 3: When considering the total claims, if hypothesis 2 holds and the claim amount {X_ij, j> 1, i ⩾ 1 } is a sequence of non-negative random variables with the same distribution F. At the same time, it satisfies hypothesis 4, and there is F ∈ C, and there is a certain $p > J_{F}^{+}$ such that ${EX}_{ij}^{2 p + 1} < \infty, i, j ⩾ 1$ , then There is γ₁ > 0 such that for any γ ⩾ γ₁, there is: $P (S (t) > x) \sim v λ_{t} \bar{F} (x)$ (24)

It is consistent for all x ⩾ γt. Equivalently, there is: $lim_{t \to \infty} sup_{x ⩾ γ t} | \frac{P (S (t) > x)}{v λ_{t} \bar{F} (x)} - 1 | = 0$ (25)

4 Internet financial trading platform model

This article takes H enterprise as an example to construct the model of the Internet agricultural financial transaction platform. The supply chain finance of the collateral model is shown in Fig. 1. The collateral mode is mainly aimed at financial solutions in the case of large purchases such as peak consumption during holidays, large-scale promotional activities, and related products in short supply.

Fig. 1

Supply chain finance of collateral model.

Compared with the collateral model, the credit model is mainly aimed at dealers’ daily purchases on a monthly basis. Therefore, the purchase amount is relatively small, the process is relatively simplified, and the capital turnover is convenient. The credit model supply chain finance is shown in Fig. 2.

Fig. 2

Supply chain finance of credit model.

The specific process of the credit model is as follows. Distributors provide pre-purchase orders for the current month to the supply chain financial system of H enterprise. Then, the smart factory of H enterprise arranges production according to the pre-purchase order. After that, the supply chain financial system of H enterprise and the commercial bank analyze the credit status and provide full funds to the H enterprise’s finance enterprise based on the credit records accumulated by the dealer. After that, the finance enterprise received the payment and allowed the factory to ship the goods. After that, the smart factory will deliver the goods to the dealers through the RI RI Shun logistics system. Finally, the dealer pays the money to the commercial bank or relevant financial institution after receiving the goods.

5 Model data processing

In the analysis of corporate financial status, profitability, solvency and operational capabilities can most comprehensively and intuitively reflect the current financial status of the company. Through the comparative analysis of corporate financial data before and after the implementation of the supply chain finance model, it can reflect the notable financial problems of the company under the supply chain finance model, as well as the hidden risks indirectly reflected from the changes in financial data. From the perspective of an enterprise, maximizing profits is the fundamental purpose of business operations, and more profits are the prerequisite for supporting enterprises to expand their scale and continue to create value. Companies with stronger profitability have more vitality and good development prospects. In order to fully reflect the level of corporate profitability, this paper selects the profitability of production and operation and the profitability of assets for detailed analysis, as shown in Table 1 and Fig. 3.

Table 1
Statistical table of profitability related data

2019/12/31 2018/12/31 2017/12/31 2016/12/31 2015/12/31 2014/12/31

Return on total assets 6.04% 5.15% 7.88% 9.01% 9.19% 8.87%

OPE 30.79% 30.75% 27.79% 27.34% 25.07% 24.95%

Net interest rate 6.46% 6.52% 7.93% 9.94% 10.13% 9.85%

Operating profit margin 6.41% 6.04% 7.26% 8.80% 7.20% 6.67%

Sales margin 5.74% 5.68% 6.67% 7.62% 6.48% 5.51%

Return on assets 12.34% 11.13% 15.95% 16.19% 17.97% 18.76%

	2019/12/31	2018/12/31	2017/12/31	2016/12/31	2015/12/31	2014/12/31
Return on total assets	6.04%	5.15%	7.88%	9.01%	9.19%	8.87%
OPE	30.79%	30.75%	27.79%	27.34%	25.07%	24.95%
Net interest rate	6.46%	6.52%	7.93%	9.94%	10.13%	9.85%
Operating profit margin	6.41%	6.04%	7.26%	8.80%	7.20%	6.67%
Sales margin	5.74%	5.68%	6.67%	7.62%	6.48%	5.51%
Return on assets	12.34%	11.13%	15.95%	16.19%	17.97%	18.76%

Fig. 3

Statistical diagram of profitability related data.

The solvency of an enterprise is an important indicator that reflects the financial status and operating ability of an enterprise, and is the key to whether an enterprise is healthy. This section analyzes corporate solvency from two perspectives: short-term solvency and long-term solvency. For short-term debt solvency, this article selects two indicators: current ratio and quick ratio. For the long-term solvency, this article selects the asset-liability ratio and the equity ratio to explain it, as shown in Table 2 and Fig. 4.

Table 2

Related data of solvency

	2019/12/31	2018/12/31	2017/12/31	2016/12/31	2015/12/31	2014/12/31
Current ratio	1.16%	0.96%	1.39%	1.44%	1.31%	1.28%
Quick ratio	0.88%	0.75%	1.17%	1.26%	1.13%	1.05%
Cash ratio	46.21%	32%	62.74%	69.50%	54.81%	52.48%
Assets and liabilities	69.82%	72.08%	57.91%	61.79%	67.90%	69.64%
Equity ratio	200.77%	239.15%	124.94%	144.38%	191.98%	205.60%

Fig. 4

Statistics chart of related data of solvency.

Due to the risks of business operations and the existence of uncertain factors, companies need to have a certain amount of current assets that can be immediately realized to repay current liabilities. Different from the current ratio, this part of current assets requires higher liquidity. According to the data from the annual report of H enterprise in Table 2, it can be seen that under the new model, enterprise inventories have decreased and liquid capitals have increased. However, affected by the supply chain finance model, current liabilities increases, and the increase in current liabilities is greater than that of liquid capitals, which led to a decline in the overall quick ratio.

When analyzing the quick ratio, the quality of accounts receivable should also be analyzed to assist in judging the enterprise’s short-term solvency. According to the quality analysis data of accounts receivable in Table 3 and Figs. 5 to 6, it can be seen that the turnover rate and turnover days of accounts receivable have not been as good as before after the transformation of H enterprise in 2016. Moreover, the “zero inventory” strategy has greatly increased the inventory turnover rate of H enterprise and shortened the turnover days. However, most upstream and downstream SMEs do not have the same capital scale as H enterprise, and cannot complete the accounts in time in the delivery and payment link. This is one of the reasons why H enterprise adopts the supply chain finance strategy. The quality analysis data of accounts receivable is shown in Chart 3, and the corresponding statistical diagrams are shown in Figs. 5 and 6.

Table 3

Analysis data of the quality of accounts receivable

	Book balance	Bad debt provision	Bad debt rate	Accounts receivable turnover rate	Accounts receivable turnover days
2019	12931519117	359034236.1	2.81%	13.029	28.1891
2018	12711033292	341316752.9	2.72%	13.0795	28.078
2017	6529268253	326463412.6	5.05%	15.8469	23.1694
2016	5630252100	281512605	5.05%	18.6345	19.7051
2015	4600109690	230005484.5	5.05%	20.4929	17.9174
2014	4461776361	223088818.1	5.05%	22.1594	16.5741

Fig. 5

Statistical diagram of book statistics amount.

Fig. 6

Statistical diagram of accounts receivable turnover rate and turnover days.

The operating capability of an enterprise is mainly reflected in the turnover efficiency of related assets, inventory and other fixed assets. Since in the traditional manufacturing industry, the index of turnover efficiency reflected by operating capacity is one of the factors of manufacturing profitability and development, operating capacity analysis is an important supplementary explanation after profitability and solvency analysis. Since the analysis of turnover rate in a single enterprise’s operating capability index is weak, this article uses the current domestic agricultural representative M Group and G Group to conduct a horizontal comparative analysis to reflect the industry position of operating capability of H enterprise in the main domestic appliance manufacturing market. After that, this article compares the values before and after 2014 to analyze the impact of supply chain finance models on the enterprise’s operating capabilities. The results are shown in Tables 4 6, and the corresponding statistical diagrams are shown in Figs. 7 9.

Table 4

Operational capability indicators om 2019.12.31

	M	G	H
Turnover rate of accounts receivable (times)	15.54	33.8	12.9
Accounts receivable turnover days (days)	23.17	10.65	27.91
Inventory turnover rate (times)	8.01	7.78	5.98
Turnover rate of fixed assets (times)	11.02	8.43	10.07
Turnover rate of total assets (times)	1.15	0.75	1.13
Inventory turnover days (days)	44.96	46.27	60.18
Total asset turnover days (days)	313.1	482.32	319.55
Liquid assets turnover rate (times)	1.66	0.94	2.02
Current asset turnover days (days)	217.18	381.68	178.41

Table 5

Operational capability indicators on 2018.12.31

	M	G	H
Turnover rate of accounts receivable (times)	13.35	37.09	12.95
Accounts receivable turnover days (days)	26.97	9.71	27.8
Inventory turnover rate (times)	8.87	7.88	6.9
Turnover rate of fixed assets (times)	7.99	6.53	9.91
Turnover rate of total assets (times)	1.06	0.63	1.15
Inventory turnover days (days)	40.6	45.69	52.16
Total asset turnover days (days)	338.89	571.88	313.26
Liquid assets turnover rate (times)	1.49	0.82	1.91
Current asset turnover days (days)	242.18	438.54	188.04

Table 6

Operational capability indicators on 2017.12.31

	M	G	H
Turnover rate of accounts receivable (times)	14.03	35.28	15.69
Accounts receivable turnover days (days)	25.66	10.2	22.94
Inventory turnover rate (times)	8.06	7.31	8.02
Turnover rate of fixed assets (times)	7.23	6.43	11.62
Turnover rate of total assets (times)	1.11	0.61	1.19
Inventory turnover days (days)	44.66	49.28	44.87
Total asset turnover days (days)	323.92	585.46	302.78
Liquid assets turnover rate (times)	1.54	0.81	1.57
Current asset turnover days (days)	233.77	443.95	229.33

Fig. 7

Statistical diagram of operating capability indicators on 2019.12.31.

Fig. 8

Statistical diagram of operating capability indicators on 2018.12.31.

Fig. 9

Statistics chart of operational capability indicators on 2017.12.31.

Based on the comparative analysis of the above three aspects of profitability, solvency, and operating ability, we found that before and after the adoption of supply chain finance in 2016, the financial indicators of H enterprise have been significantly optimized and improved in terms of the profit rate of main business and the asset-liability ratio. It shows that the supply chain finance model has promoted the development of the overall supply chain, making production and sales more targeted, and the sufficient funds of downstream small and medium dealers have also promoted product ordering and sales, and increased the profit income of core enterprises and downstream distributors. At the same time, the development of the new model of core enterprises is also a process of constantly urging enterprises to optimize their capital structure. Only a reasonable capital structure can ensure the smooth flow of funds in the production and marketing process and ensure that core enterprises will not be affected by accidental capital turbulence. However, at the same time, the supply chain finance model has also brought a certain impact on core companies in terms of cost control, account receivable and inventory turnover efficiency, and long-term debt solvency. The turnover efficiency of accounts receivable and inventory restricts the return of funds, which ultimately affects the long-term solvency of enterprises and exposes core enterprises to capital risks.

Through the above introduction to the financial process of the H enterprise Internet supply chain and the risk identification and control measures, it can be found that the supply chain financial model involves obvious characteristics of the entire industry structure. In order to carry out risk control more systematically and comprehensively, H enterprise should expand the financial layout of the supply chain, truly control all links in the supply chain process, and realize a digital management system from prevention to monitoring. H enterprise also established its own logistics enterprise in the logistics link, which not only provides transportation support for its own product production and sales, but also undertakes order collection and product feedback. This is similar to the production and sales model of similar products in the same industry, and it is difficult to develop innovative advantages. At the same time, the risk of the supply chain finance model under the Internet still lies in its uncontrollability. The imbalance of production and distribution is the main source of the supply chain financial capital break and the credit crisis. In response to this, H enterprise should continue to optimize the advantages of the Internet, and at the same time extend the scope of control to various local distribution agencies, realize sales networking, and timely feedback on order sales performance. Moreover, it needs to upload monthly or quarterly sales data together and compare them together. In addition, it needs to verify whether the dealer’s planned sales volume has a large error with the actual sales volume, whether the error occurs, and whether the inventory is capable of continuing to sell. After the normal operation analysis, the head office will conduct systematic guidance. If continuous abnormal business conditions occur, the enterprise needs to conduct on-site investigations in time to eliminate risks. For manufacturing companies, sales are the source of profits, and sales performance is the most powerful evidence that reflects whether the operating conditions are good. Large-scale manufacturing companies have established an ecological environment and have incorporated distributors into the supply chain model. Therefore, the management of dealers should be continuously strengthened, and the control of potential risks of dealers is also the key to the control of financial risks in the Internet supply chain.

6 Conclusion

Internet finance supports the development of agricultural industrialization, has played a huge role in promoting the development of inclusive finance in rural areas and increasing the coverage of small and micro enterprises in agriculture, which has led to a new explosive growth in the entire agricultural industry. This article combines Internet finance with the development of agricultural industrialization, and uses agricultural industrialization theory and Internet finance theory to study the impact of Internet financial support on the development of agricultural industrialization in my country. According to the current main problems of Internet finance supporting agricultural development, this paper proposes a policy support system for improving Internet finance to support the development of agricultural industrialization. Moreover, this article puts forward policy recommendations such as vigorously developing agricultural e-commerce finance, expanding mobile finance’s coverage of agriculture, and improving the Internet financial support and supervision system for agricultural industrialization. We have reason to believe that the “Internet + finance + agriculture” model will fundamentally solve the funding needs of my country’s agricultural development, allowing more investment to see the prospects of investment in the agricultural industry. Moreover, it can help more agricultural enterprises to carry out “Internet +” deep-level reforms to promote the process of my country’s agricultural intelligence.

Footnotes

Acknowledgments

Major Project of Sichuan Federation of Social Science Associations: Research on the formation mechanism and realization path of green production behavior of farmers’ professional cooperatives(SC19A009).

References

Cole

E.S.

, Walker

, Mora

, et al., Identifying hospitals that may be at most financial risk from Medicaid disproportionate-share hospital payment cuts, Health Affairs 33(11) (2014), 2025–2033.

Zalik

, Resource sterilization: reserve replacement, financial risk, and environmental review in Canada’s tar sands, Environment and Planning A 47(12) (2015), 2446–2464.

Lucarelli

, Uberti

and Brigheti

, Misclassifications in financial risk tolerance, Journal of Risk Research 18(4) (2015), 467–482.

Chan

E.Y.

, Physically-attractive males increase men’s financial risk-taking, Evolution and Human Behavior 36(5) (2015), 407–413.

West

and Worthington

A.C.

, Macroeconomic conditions and Australian financial risk attitudes, 2001–2010, Journal of Family and Economic Issues 35(2) (2014), 263–277.

Tanimura

, Jaramillo

, Weil

, et al., Financial burden for tuberculosis patients in low-and middle-income countries: a systematic review, European Respiratory Journal 43(6) (2014), 1763–1775.

Scharding

T.K.

, Imprudence and immorality: A Kantian approach to the ethics of financial risk, Business Ethics Quarterly 25(2) (2015), 243–265.

and Furman

, A form of multivariate Pareto distribution with applications to financial risk measurement, ASTIN Bulletin: The Journal of the IAA 47(1) (2017), 331–357.

Wang

G.W.Y.

, Woo

and Mileski

, The relative efficiency and financial risk assessment of shipping companies, Maritime Policy & Management 41(7) (2014), 651–666.

10.

Luck

L.J.

, Justice and systemic financial risk, Journal of Applied Philosophy 34(3) (2017), 331–352.

11.

Allen

and Petsoulas

, Pricing in the English NHS quasi market: a national study of the allocation of financial risk through contracts, Public Money & Management 36(5) (2016), 341–348.

12.

Thomä

and Chenet

, Transition risks and market failure: a theoretical discourse on why financial models and economic agents may misprice risk related to the transition to a low-carbon economy, Journal of Sustainable Finance & Investment 7(1) (2017), 82–98.

13.

Sosnovska

and Zhytar

, Financial architecture as the base of the financial safety of the enterprise, Baltic Journal of Economic Studies 4(4) (2018), 334–340.

14.

Acemoglu

, Ozdaglar

and Tahbaz-Salehi

, Systemic risk and stability in financial networks, American Economic Review 105(2) (2015), 564–608.

15.

Castellano

, Cerqueti

and Rotundo

, Exploring the financial risk of a temperature index: A fractional integrated approach, Annals of Operations Research 284(1) (2020), 225–242.

16.

Cohn

, Engelmann

, Fehr

, et al., Evidence for countercyclical risk aversion: An experiment with financial professionals, American Economic Review 105(2) (2015), 860–885.

17.

Maggiori

, Financial intermediation, international risk sharing, and reserve currencies, American Economic Review 107(10) (2017), 3038–3071.

18.

Pinelis

, An optimal three-way stable and monotonic spectrum of bounds on quantiles: A of coherent measures of financial risk and economic inequality, Risks 2(3) (2014), 349–392.

19.

Hautsch

, Schaumburg

and Schienle

, Financial network systemic risk contributions, Review of Finance 19(2) (2015), 685–738.

20.

Noussair

C.N.

, Trautmann

S.T.

and Van de Kuilen

, Higher order risk attitudes, demographics, and financial decisions, Review of Economic Studies 81(1) (2014), 325–355.

21.

Rodríguez-Aguilar

, Marmolejo-Saucedo

J.A.

and Tavera-Martínez

, Financial risk of increasing the follow-up period of breast cancer treatment currently covered by the Social Protection System in Health in Mexico, Cost Effectiveness and Resource Allocation 16(1) (2018), 1–11.

22.

Lockwood

, Predicting the unpredictable: Value-at-risk, performativity, and the politics of financial uncertainty, Review of International Political Economy 22(4) (2015), 719–756.

23.

Gallardo Millán

L.A.

and Ochoa

E.A.

, Estructura De Capital Y Riesgo Financiero: Evidencia Empírica En Pymes Hoteleras (Capital Structure and Financial Risk: Empirical Evidence in SMEs), Revista Global de Negocios 3(4) (2015), 1–10.

24.

Cueva

, Roberts

R.E.

, Spencer

, et al., Cortisol and testosterone increase financial risk taking and may destabilize markets, Scientific Reports 5(1) (2015), 1–16.

25.

Lee

C.J.

and Andrade

E.B.

, Fear, excitement, and financial risk-taking, Cognition and Emotion 29(1) (2015), 178–187.