STUDY ON THE INFLUENCING FACTORS

AND IMPROVING COUNTERMEASURES OF

REGIONAL FINANCIAL SERVICE FUNCTION

- TAKE CORPORATE LOANS AS AN

EXAMPLE

Suo Zhang

• Graduate School, Jose Rizal University, Manila, 0900, Philippines

• School of Intelligent Manufacturing and Mechanical Engineering, Hunan

Institute of Technology, Hengyang, Hunan, 421002, China

Yixian Wen*

• School of Business, Hunan Institute of Technology, Hengyang, Hunan, 421002,

China

•E-mail: wenyixian@hnit.edu.cn

Reception: 27 December 2023 | Acceptance: 22 January 2024 | Publication: 21 February 2024

Suggested citation:

Zhang, S. and Wen, Y. (2024) Study on the inﬂuencing factors and

improving countermeasures of regional ﬁnancial service function -- Take

corporate loans as an example. 3C TIC. Cuadernos de desarrollo aplicados a

las TIC, 13(1), 62-74. https://doi.org/10.17993/3ctic.2024.131.62-74

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

ABSTRACT

Focusing on regional financial services, this paper analyzes the variables affecting

non-financial businesses getting loans through the region economic services function,

and then proposes countermeasures to improve it. The DPC algorithm is applied to

the financial analysis in this paper because of its fast speed and high accuracy. The

AD-DPC approach is suggested in this study as a solution to the issue that the

computation of local density relies on the choice of the truncation length parameter

and the clustering sites must be manually chosen. This strategy lessens the

subjectivity and volatility that the fictitious label

brings. For the DPC algorithm by

using a one-step assignment strategy, i.e., assigning the labels of clustering centers to

all non-clustering centroids at one time, such a strategy is poorly fault-tolerant, this

paper proposes the DAS-DPC algorithm on the basis of AD-DPC. Through

experiments, ADAS-DPC is optimal for ARI metrics in the dataset. Among them, the

ARI indexes of ADAS-DPC algorithm are 0.832, 0.895, 0.768 and 0.757 in the

datasets Iris, Wine, Seed and Sonar. It shows that the ADAS-DPC algorithm can not

only handle the datasets with complex shapes, large density differences between

clusters and tightly connected clusters, but also improve the clustering performance of

the algorithm for high-dimensional data.

KEYWORDS

Regional financial services; corporate lending; DPC algorithm; AD-DPC algorithm;

DAS-DPC algorithm

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

INDEX

ABSTRACT .....................................................................................................................2

KEYWORDS ...................................................................................................................2

1. INTRODUCTION .......................................................................................................4

2. FINANCIAL ANALYSIS MODEL BASED ON IMPROVED DPC ALGORITHM .......5

2.1. Adaptive density peak clustering algorithm ........................................................5

2.1.1. Improved local density calculation ...............................................................5

2.1.2. Finding the optimal neighborhood size ........................................................6

2.1.3. Adaptive determination of clustering centers ..............................................7

2.2. Adaptive multi-step allocation strategy density peak clustering algorithm .........8

2.2.1. Non-core point division ................................................................................9

2.2.2. Label assignment ......................................................................................10

2.3. Performance analysis of the algorithm .............................................................12

3. STUDY ON THE IMPACT FACTORS OF REGIONAL FINANCIAL SERVICES

FUNCTION ..............................................................................................................14

3.1. Research Program ...........................................................................................14

3.1.1. Research hypothesis .................................................................................14

3.1.2. Research Methodology .............................................................................15

3.1.3. Data sources .............................................................................................15

3.2. Econometric tests of factors influencing corporate lending ..............................16

4. SUGGESTIONS FOR COUNTERMEASURES ......................................................19

5. CONCLUSION ........................................................................................................20

ABOUT THE AUTHORS ...............................................................................................22

FUNDING ......................................................................................................................22

REFERENCES ..............................................................................................................22

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

1. INTRODUCTION

The primary engine of economic growth is finance, and the "vitality", "stability",

"prosperity" and "strength" of finance determine the "vitality", "stability", "prosperity"

and "strength" of the economy. Determines the "vitality", "stability", "prosperity", and

"strength" of the economy, pointing up the crucial role that money plays in economic

growth [1-3]. The most potent indicator of increased financial competitiveness is the

improvement in the capacity of financial services to contribute to the actual economy,

the development of inclusive finance, or the promotion of an equitable growth of the

capital market [4]. Regional financial competitiveness, on the other hand, is the

expression of financial competitiveness at the regional level, as well as the relative

advantage and comprehensive ability of a region in competition with other regions

through the process of absorption, control, utilization, ownership and allocation of

financial resources [5-6]. Therefore, the level of a region's financial competitiveness

directly determines whether the region's financial resources are effectively allocated, it

ultimately has an impact on the area's capacity to deliver financial goods to the actual

economy, thereby affecting the overall national economic and social development

[7-9].

The literature [10] specifically analyzes the causes of regional financial differences

and the impact of diffusion methods on regional financial development differences,

and finds that financial liberalization has a key impact on the unbalanced development

of regional finance, and this unbalance leads to the emergence of administrative

power inequality, which makes financial development differences more serious, and a

vicious circle is thus formed, and the differences gradually expand. The literature [11]

argues that credibility in the financial market has an important influence on the

transaction behavior of borrowers and lenders, just as appearance has a positive

effect on a person's employment, good credibility has a positive effect on access to

loans, and good credit and high market recognition make it easier to obtain loans. The

literature [12] suggests that blockchain technology can be applied to supply chain

finance projects, which is expected to accelerate the speed of capital operation in the

whole supply chain platform, but it is still in the exploration stage. According to the

literature [13], while increasing financial growth scale does not significantly affect the

outcomes, improving financial growth structure and efficiency are advantageous to

enhancing regional innovation potential. The literature [14] found through the study of

Internet finance that the relationship between traditional finance and Internet finance

is one of competition and integration, and the competitiveness of Internet finance is

increasing. By studying the synergy between market-oriented financial

competitiveness and government-oriented financial stability, the literature [15] found

that there is a high correlation between financial competitiveness and financial

stability, and the dispersion of their synergistic effects and economic development are

strongly correlated.

The DPC clustering algorithm is known for its simplicity and efficiency, but there are

some drawbacks. To address the problem that the calculation of local density depends

on the selection of the truncation distance parameter

and the clustering centers

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

need to be selected manually, this paper proposes an adaptive local density and

clustering center density peaking algorithm (AD-DPC). The algorithm finds the optimal

number of nearest neighbors through the concepts of "forward neighbors" and

"reverse neighbors", and proposes a adjacent neighbor fuzzy kernel to determine

the regional density. This approach reduces the subjectivity and instability caused by

artificially specified . In this paper, based on the multiplication of local density and

relative offset distance, a new core point evaluation method is proposed, through

which core points can be effectively screened. For the DPC algorithm by using a one-

step assignment strategy, i.e., assigning the labels of clustering centers to all non-

clustering centroids at one time, such a strategy is poorly fault-tolerant, this paper

proposes an adaptive multi-step assignment strategy for density peak clustering

algorithm (ADAS-DPC) on the basis of AD-DPC. The three primary steps of the

suggested method are as follows: the first step employs a particular strategy to locate

edge points and high-density points. In the second step, the core point labels are

assigned by propagating the assignment strategy. In the third step, a checking

mechanism is introduced to check whether the labels of the edge points are optimal.

Finally, the ADAS-DPC algorithm is applied to the analysis of factors influencing the

function of regional financial services.

2. FINANCIAL ANALYSIS MODEL BASED ON

IMPROVED DPC ALGORITHM

2.1. ADAPTIVE DENSITY PEAK CLUSTERING ALGORITHM

In the DPC algorithm, if the truncation distance is not chosen properly, that could

have a significant impact on how the regional density of points is calculated and

ultimately result in a reduction in the clustering effect, while in this section, an

improved local density calculation will be proposed without artificially coming to set the

truncation distance. Based on the data provided by its nearby samples, the regional

density of a site is precisely determined.

2.1.1. IMPROVED LOCAL DENSITY CALCULATION

This section suggests using a fuzzy kernel to estimate local density as a better

method of doing so. The idea of nearest neighbor and nearest neighbor will be

introduced by the suggested fuzzy kernel in this section in order to take the local data

structure into account while determining the densities. As described by the proposed

fuzzy kernel:

(1)

i= max 1 −1

k∑

j∈knn(xi)

d(xi,xj),0

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

where is the group of close neighbors of point . It's described as:

(2)

where is the Euclidean distance among and and is the th nearest

neighbor of . The information on the local density dispersion from nearby samples is

combined in the enhanced local density expression. As a result, the fuzzy kernel can

effectively extract the local density distribution. The focus on neighbor relationships is

appropriate for sets of data with an uneven density dispersion and subpar DPC

performance.

2.1.2. FINDING THE OPTIMAL NEIGHBORHOOD SIZE

In the previous section, a way to calculate the local density using

nearest

neighbor was proposed, where the parameter still needs to be set artificially, which

also interrupts and destroys the continuous operation of the algorithm to some extent.

In this section, the optimal value is determined by calculation, so that the whole

algorithm does not depend on the parameter.

In the present study, we suggest a flexible method for finding the optimal value.

Firstly, by setting the value initially to 1, the nearest neighbor point of each point is

found by definition 1, and then the number of times each point is considered as a

neighbor point by other points, i.e., the reverse neighbor proposed by definition 2, is

calculated to help iterative finding. Then the optimal

value is finally found by

increasing in steps of 1.

The Euclidean distance among two data points and of dimension

represented by . and fit to the set , is the th neighbor of , and is

sorted by distance from , from smallest to largest.

Definition 1 Positive neighbors and the positive neighbors of point

are

represented by set :

(3)

Definition 2 Reverse neighbor, the reverse neighbor point of any point

represented by the set , which needs to satisfy equation (4), i.e., a point in

the forward neighbor of point . If the forward neighbor point of contains point ,

then is said to be a reverse neighbor. is used to denote the set of reverse

neighbor points of point :

(4)

knn(xi)

nn (xi)=

{

xj∣d

(

xi,xj

)

≤d(xi,xk)

}

d(xi,xj)

Dis(xi,xj)

nbr(xi)

br (xi)=

{

xm∣Dis (xi,xm)< Dis

(

xi,xj

)}

r nbr (xi)

rnbr (xi)={xn∣xi∈nbr (xn)∧xn∈nbr (xi)}

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

The optimal value size can be obtained as follows: as the value increases from

1 and in increments of 1, each point in the set acquires at least one reverse

neighbor point, but when all points do not acquire a reverse neighbor point, The

value is regarded as the ideal value. The termination condition for the optimal

value is expressed using equation (5):

(5)

Where, starts from 1 with a step size of 1 and sets the upper search limit

. indicates the number of elements of the set of

. Combining with equation (1), the local density can be obtained.

2.1.3. ADAPTIVE DETERMINATION OF CLUSTERING CENTERS

The DPC algorithm requires visual inspection of the decision diagram to manually

set the clustering centers. This human selection approach is unreliable when dealing

with complex decision graphs. In order to select the appropriate core points, this

section will design a new scoring formula to evaluate each point and then check

whether it can be considered as a clustering center based on the threshold value. In

this paper, the improved local density and relative offset distance will be used to

identify the clustering centers. Then a new scoring method is proposed for scoring all

points to determine the clustering centers by scoring, see Equation (6).

(6)

Using the assessment score values , each point in the data set is sorted.

Higher score values will be assigned to points that have elevated local density and

high comparative offset distances, and then the scores are sorted in descending order

from highest to lowest. In order to determine the candidate clustering centers a

threshold value is needed, by which the clustering centers can be determined

adaptively.

For the proposed design algorithm finds the threshold value. As shown in

Figure 1, the values of for most of the data points are concentrated in a

lower region, while the values of for only a few points are concentrated in a

higher region. The non-clustering center point of decreases almost linearly

and slowly. From the critical point value to the non-clustered centroid

value, there is a leap gap. The core points are filtered by finding this gap.

Firstly, this subsection defines equation (7) to describe the absolute

(xi)=

r nbr

(

)

−r nbr

(

)

k−1

r nbr (xi)k

l(l= int( n))

r nbr

(

)

r nbr (xi)

Pscore

(ρ

max(ρ) max(δ))2

Pscorei

Pscore

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

difference of adjacent consecutive points, and then calculates the mean value of

by equation (8).

(7)

(8)

In Figure 1(a), the solid blue circles represent the core points and the hollow blue

circles represent the non-core, and from Figure 1(b) the vertical coordinates are the

scoring values and the horizontal coordinates are the point numbers. We can see that

the points behind the core points are very small. This feature can be used

to determine the candidate core points.

Figure 1. Data distribution and value ranking chart

When there are multiple candidate centers in a high-density region, they are

usually very close to each other. Therefore, it is necessary to determine whether these

candidate centers can be identified as the final independent clustering centers.

Therefore, to determine each candidate point's closest neighbor, the candidate

cores are adjusted. If, for a point in the candidate set, a nearest neighbor is

found and is also a candidate core, then the values of the Pscore of the two points

are compared and the one having the larger Pscore value is kept and the smaller one

is removed from the candidate core set. The final actual clustering centers are

determined from the candidate set using a fine-tuning process.

2.2. ADAPTIVE MULTI-STEP ALLOCATION STRATEGY DENSITY

PEAK CLUSTERING ALGORITHM

Although the AD-DPC algorithm proposed above solves the two problems that the

local density calculation in the DPC algorithm depends on the truncation distance and

requires human experience to select the clustering centers based on the decision

map. When the data set is vast, there are many noisy points, and the area of overlap

is complicated, the accuracy index still deteriorates. To address this problem, the

DPscore

DPscoreei=abs(Pscorei−Pscorei+1)

Pscore =∑

i∈D

DPscore

|D|

−1

DPscore

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

adaptive multi-step assignment strategy density peak clustering algorithm, referred to

as the ADAS-DPC algorithm, will be proposed on the basis of the AD-DPC algorithm.

2.2.1. NON-CORE POINT DIVISION

The main purpose of this step is to make an initial distinction between non-core,

and to first perform a round of division of non-core points by the formula designed in

this subsection. This division is not a division of non-core point categories, but a round

of data points based on the distribution in the decision diagram. For this purpose,

three types of points are divided: cluster centers, high-density points, and edge points.

While the density near edge points differs significantly from that of their neighbors, it is

similar to that of the high-density points and their neighbors. Depending on how

effectively the clustering algorithm interprets the clustering structure, edge point

detection accuracy can vary. The sites with substantial density values created by

eliminating edge points are the clustering backbones. Each core should roughly

preserve the cluster's form and be separated from the others:

(9)

where is the number of points.

Edge points are points surrounding the backbone cluster, so these points have

different characteristics from those in the high-density cluster. To express this concept,

edge points are points whose local density is lower than the average

local density, and their relative offset distance values are lower than the variance

of the relative offset distance values of all points, and the effect of

setting is to exclude individual points with abnormal values, see Equation (10):

(10)

Figure 2 shows the distribution of data points in the decision diagram, with the

horizontal coordinates representing the local density and the vertical coordinates

representing the relative offset distance. The blue and orange colors represent the

core points, the black circles represent the high density points, and the black

diamonds represent the edge points. This figure clearly shows the distribution of core,

high-density, and edge points in the decision diagram. divides the diagram into two

parts, left and right, with core and high-density points on the right side of the diagram

and edge points on the left side of the diagram.

Avg

(ρ) =

∑

i=1

(ρi<Avg(ρ))

(δi<Var(δ))

Var(δ)

Var

(δ) =

∑

i=1

(δi−¯

δ)

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

Figure 2. Distribution of data points in the decision diagram

2.2.2. LABEL ASSIGNMENT

This subsection aims to assign clustering labels to non-clustering centroids by three

sub-steps: (1) Core points are assigned to high-density points. (2) High-density points

are assigned to high-density points. (3) High-density points are assigned to edge

points.

The cluster centers will each be given a different label, after which the tags of the

comment cluster centers will be discovered by searching for the cluster centers'

mutually dense nearest neighbors. The marks with all these assigned labels will then

be used as starting points, and each nearest neighbor will then be found dynamically.

This concept is expressed by equation (11), where denotes the label of .

(11)

Figure 3 illustrates the specifics of the suggested label assignment technique for

arbitrary data. Figure 3(a) highlights that the clustering centers have been identified

using the AD-DPC algorithm proposed in this paper. The proposed approach in this

chapter is then used to identify high-density points and border points, with high-

density points highlighted by hollow black circles and black hollow diamonds

indicating edge points. The tags of a cluster centers are transmitted to their dense

neighbors, as shown in Figure 3(b). The left cluster's trunk points are depicted in this

figure as blue, as well as the right cluster's trunk points as orange.

The second stage in this work tries to transmit the clustering tags to the edges by

assigning the tags of the closest trunk points in order to decrease computation time.

L abel(xi)

Label

(xi)=

{

Label

(

)

xi∈r nbr

(

)

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

Equation (12) is used to determine the total of each edge point's distances from all of

its backbones' nearest neighbors for this purpose:

(12)

where is an edge point, is a collection of dense nodes from the rd trunk,

and depicts the distance in Euclidean terms between the two spots.

is the average of all the distances between and the closest

neighbors of in . The edge point will obtain the trunk label with the smallest

distance from itself. The purpose of using the exponential function in Eq. (13) is to

magnify the gap and avoid the inability to distinguish the size due to insufficient

number of calculated bits when the distances are extremely close:

(13)

where is the total number of detected trunks. This method is repeated until

cluster labels have been assigned to each edge point. The proposed method's

assignment of the labels to edge points is depicted in Figure 3(c). The figure

demonstrates that point has been given by the blue label using the method

described in this study because the average amount of the distances added by the

blue labels is less than the average value of the distances added by the orange

labels.

Figure 3(d) shows the edge point rechecking mechanism, which starts to execute

after the labels of all edge points have been assigned, to detect whether the edge

point label assignment is reasonable.

Figure 3. Example of a multi-step allocation strategy

SumDis (

xb,Mc

)

k∑

i∈

nbr

(

)

xb−xi

−x

SumDis (xb,Mc)

Label (x

)= argmin

c∈C

eSumDis(x

,Mc)

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

2.3. PERFORMANCE ANALYSIS OF THE ALGORITHM

To verify the effectiveness of the algorithm, comparative experiments are

conducted using UCI datasets, which contain a wide variety of sizes and types. Some

of these datasets also have clusters of various complex shapes, and the data

distribution of the specific UCI datasets is exposed in Table 1. It is visible from the

table that these seven UCI datasets basically cover a wide variety of datasets (size,

complexity and data characteristics). The performance and robustness of the

algorithm can be detected by such a variety of types of data with complex distribution.

This section tests the performance of the ADAS-DPC, AD-DPC, and DPC algorithms

by using the evaluation metrics FMI and ARI, showing the best values for each test

dataset.

Table 1. UCI data set

Figure 4 shows the results of FMI for the clustering algorithm on the seven

datasets. From the overall results the ADAS-DPC algorithm has the highest FMI

metrics in the datasets. Among them, in the datasets Iris, Wine, Seed, Sonar, and

WDBC, the FMI values of ADAS-DPC are all over 0.8, which shows that the ADAS-

DPC algorithm is effective for the improvement of the AD-DPC algorithm proposed in

this paper, especially when facing the datasets with larger dimensionality, ADAS-DPC

has a certain degree of improvement compared with the algorithms participating in the

comparison experiments. Degree of improvement.

Data set Number of data Dimensionality Number of classes

Inis 150 4 3

Wine 178 13 3

Seed 210 7 3

Sonar 208 60 2

Ecoli 336 8 8

WDBC 569 30 2

Balance 625 4 3

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

Figure 4 Experimental results of FMI in UCI dataset

Figure 5 shows the results of the clustering algorithm for the ARI metrics in the

seven datasets. The overall results show that ADAS-DPC has the best ARI metrics in

the datasets. Among them, in the datasets Iris, Wine, Seed, and Sonar, the ARI

metrics of the ADAS-DPC algorithm are 0.832, 0.895, 0.768, and 0.757, which shows

that the ADAS-DPC algorithm is effective for the improvement of the AD-DPC

algorithm proposed in this paper. In the Ecoli, WDBC, and Balance datasets, although

ADAS-DPC achieves the optimum, the ARI metrics are too low, indicating that it is

difficult for these eight algorithms to deal with these three datasets effectively.

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

Figure 5. ARI in the UCI dataset Experimental results

3. STUDY ON THE IMPACT FACTORS OF REGIONAL

FINANCIAL SERVICES FUNCTION

3.1. RESEARCH PROGRAM

3.1.1. RESEARCH HYPOTHESIS

As Internet finance and the actual economy grow, so do the funding options

available to businesses. However, whether it is a bank or other financing platform,

they all play the role of intermediaries for enterprise financing, collecting and lending

funds from multiple and scattered sources, Consequently, it may be said that one of

the key variables affecting how much money is lent out is the number of deposits. The

reserve minimum ratio stated by the bank's governor indirectly influences the total

quantity of loans in alongside the total number of deposits. Based on this, the

following hypotheses are put out in this article, which takes into account non-financial

company deposits, household deposits, and the required reserve ratio as variables

affecting the lending amount provided by non-financial corporations and institutions.

Hypothesis 1: The volume of loans made by non-financial institutions and

corporations has a positive correlation with the deposits made by households and

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

non-financial businesses. The ratio of reserve requirements to loans made to non-

financial institutions and corporations has a negative relationship with both variables.

Hypothesis 2: Loans to non-financial institutions and firms and loans to households

have a favorable correlation.

Hypothesis 3: The number of lending to non-financial businesses and institutions is

inversely connected with the reduction of loss-making businesses.

Hypothesis 4: Index of consumer prices Loans made to organizations and

businesses that are not in the financial sector have a negative correlation with the

CPI. Negative correlation exists between the quantity of loans placed by non-financial

firms and organizations and the Shanghai Interbank Offered Rate (SHIBOR).

3.1.2. RESEARCH METHODOLOGY

This study expanded the research area of the number of loans positioned by non-

financial businesses to the research topic of the number of loans positioned by non-

financial businesses and institutional groups by using the number of loans positioned

by non-financial businesses and governmental groups as the predictor variables.

Table 2 displays the variables.

Table 2. Selection of variables

3.1.3. DATA SOURCES

Based on the proposed research question, this paper chooses the following

variables for the period of 2017 to 2021 in Hunan Province: non-financial corporate

and institutional loan placement, non-financial corporate deposits, household

deposits, household loans, loss-making enterprises reduction rate, CPI, SHIBOR 1-

month interest rate, and deposit reserve ratio. Based on the numerous interest rate

Variable Code Variable name

YAmount of loans to non-financial enterprises

and institutions

X1 Deposits in non-financial enterprises

X2 Household deposit

X3 Household loan

X4 Reduction rate of loss-making enterprises

X5 CPI

X6 SHIBOR rate

X7 Deposit reserve ratio

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

term varieties published by SHIBOR, this paper selects the interest rate term variety

for SHIBOR term variety with the 1-month interest rate from January 2017 to

December 2021 as the choice of interest rate term variety. Since the 1-month

SHIBOR rate changes every few trading days with the market and for the

convenience of data selection, this paper selects the official January SHIBOR rate on

the first trading day of each month as the 1-month SHIBOR rate data. Regarding the

data itself, in order to further investigate the effect that outside influences of non-

financial enterprise loans on enterprise loans, this paper selects data of loan

placement of non-financial businesses and related groups in Hunan Province over the

last five years with the variables of non-financial entrepreneurship deposits and

household deposits.

3.2. ECONOMETRIC TESTS OF FACTORS INFLUENCING

CORPORATE LENDING

The results of the ADAS-DPC algorithm test were obtained using the amount of

loans to non-financial businesses and organizations as the variable that is dependent

and variables like funds to non-financial businesses and funds to households as

separate variables. The information is shown in Table 3.

Table 3's results make it clear that X1 is unimportant and X7 has a positive

correlation with Y, which is inconsistent with the hypothesis. As a result, the variables

are examined for multicollinearity, which can produce unimportant variables and the

incorrect sign of the coefficient of regression.

Table 3. ADAS-DPC estimation results

Table 4 displays the straightforward correlation coefficient matrix. Variable X7

exhibits strong multicollinearity amongst the variables since it is substantially linked

with variables X2, X3, and X4 in a two-by-two fashion. Therefore, despite the fact that

the number of loans placed and the X7 deposit reserve ratio are theoretically

Variables coefficient Standard

deviation T-value P-value

C-1.305 2.344 -553 558

X1 116 90 1.660 112

X2 281 57 3.765 -12

X3 1.005 92 11.237 3

X4 -3.194 8.344 -3.790 -1

X5 -1.016 4.636 -2.180 22

X6 -2.954 9.178 -3.228 14

X7 2.579 1.150 2.230 26

https://doi.org/10.17993/3ctic.2024.131.62-74

3C TIC. Cuadernos de desarrollo aplicados a las TIC. ISSN: 2254-6529

Ed.44 | Iss.13 | N.1 January - March 2024

negatively connected, as the deposit reserve ratio rises, banks have comparatively

less money to use for lending.

Table 4. Simple correlation coefficient matrix

On the basis of this, after deleting the variable for the X7 deposit reserve

percentage, the ADAS-DPC estimation was re-run to obtain Table 5. For the original

hypothesis Ho: , given the significance level , F(6, 53)=6.48, F=211>6.48 in the table,

the original hypothesis Ho should be rejected, demonstrating the significance of the

regression equation. Also, it can be inferred that the factors X1, X2, X3, X4, X5, and

X6 are significant at the x=5% significance threshold from the p-values in Table 5 that

are less than 0.05, showing that the regress of the chosen explanatory factors with the

explanatory variables is significant. Among them, non-financial corporations and

organizations are positively correlated with non-financial corporate deposits,

household deposits and household loans, CPI and SHIBOR 1-month interest rate are

inversely connected with the rate of loss-making businesses.

Table 5. ADAS-DPC estimation consequences

This led to an analysis of the interconnections between the variables, which is

depicted in Table 6. The regression results of the explanatory variables X2 and X3

passed the test statistics in Table 4 of the estimation findings, and there is a strong

1.000

587

576

-314

-75

109

-367

587

1.000

993

-870

225

-507

-928

576

993

1.000

-825

251

-550

-945

-314

-870

-825

1.000

-439

658

861

-75

225

251

-439

1.000

-334

-330

109

-507

-550

658

-334

1.000

703

-367

-928

-945

861

-330

703

1.000

Variables

Coefficient

Standard

deviation

T-value

P-value

3.479

9.815

3.559

252

3.493

214

3.134

-9

896

11.578

-7

-3.712

8.33

-4.478

-5

-1.46

4.357

-3.35

-1.88

8.096

-2.341