## Zhang Cheng*## |

Variety | Stage | Meaning |
---|---|---|

V1 | Cultivation stage | Agricultural mechanization services to reduce the influence of agricultural cost in the wheat cultivation stage |

V2 | Sowing stage | Agricultural mechanization services to reduce the influence of agricultural cost in the wheat sowing stage |

V3 | Plant protection stage | Agricultural mechanization services to reduce the influence of agricultural cost in the plant protection stage |

V4 | Irrigation stage | Agricultural mechanization services to reduce the influence of agricultural cost in the irrigation stage |

V5 | Fertilization stage | Agricultural mechanization services to reduce the influence of agricultural cost in the fertilization stage |

V6 | Harvest stage | Agricultural mechanization services to reduce the influence of agricultural cost in the harvest stage |

V7 | Total cost | Agricultural mechanization services to reduce the influence of the total agricultural cost |

Table 2.

*Correlation is significant at the 0.05 level (two-tailed), **correlation is significant at the 0.01 level.

V1 | V2 | V3 | V4 | V5 | V6 | V7 | ||
---|---|---|---|---|---|---|---|---|

V1 | Correlation | 1 | 1.000** | 0.337** | -0.035 | 0.288* | 1.000** | 1.000** |

Sig. (two-tailed) | 0.000 | 0.003 | 0.758 | 0.010 | 0.000 | 0.000 | ||

V2 | Correlation | 1.000** | 1 | 0.337** | -0.035 | 0.288* | 1.000** | 1.000** |

Sig. (two-tailed) | 0.000 | 0.003 | 0.758 | 0.010 | 0.000 | 0.000 | ||

V3 | Correlation | 0.337** | 0.337** | 1 | 0.919** | 0.056 | 0.337** | 0.337** |

Sig. (two-tailed) | 0.003 | 0.003 | 0.000 | 0.629 | 0.003 | 0.003 | ||

V4 | Correlation | -0.035 | -0.035 | 0.919** | 1 | -0.054 | -0.035 | -0.035 |

Sig. (two-tailed) | 0.758 | 0.758 | 0.000 | 0.640 | 0.758 | 0.758 | ||

V5 | Correlation | 0.288* | 0.288* | 0.056 | -0.054 | 1 | 0.288* | 0.288* |

Sig. (two-tailed) | 0.010 | 0.010 | 0.629 | 0.640 | 0.010 | 0.010 | ||

V6 | Correlation | 1.000** | 1.000** | 0.337** | -0.035 | 0.288* | 1 | 1.000** |

Sig. (two-tailed) | 0.000 | 0.000 | 0.003 | 0.758 | 0.010 | 0.000 | ||

V7 | Correlation | 1.000** | 1.000** | 0.337** | -0.035 | 0.288* | 1.000** | 1 |

Sig. (two-tailed) | 0.000 | 0.000 | 0.003 | 0.758 | 0.010 | 0.000 |

We transpose the original data of 78 visits and surveys in Henan Province in EXCEL, import the original data into SPSS software, and perform multiple linear regression analysis through SPSS software (IBM, Armonk, NY, USA). Therefore, the following analysis results are obtained.

Table 3 shows the linear regression analysis results of the importance and cost of agricultural mechanization services at different wheat planting stages in Henan Province. In the table, we can obtain an adjusted R2 coefficient of 1, indicating that the agricultural mechanization service levels of planting stages of harvesting stage V6, irrigation stage V4, and plant protection stage V3 can fully explain the changes in wheat cost in Henan Province. The standard error of the dependent variable’s predicted value is 0.04035, which indicates that the error between the actual observation and the regression estimate is minimal. Durbin-Watson is 1.855, close to about 2, indicating that the overall fit of the regression model is good.

Table 3.

^{a} Predictors: VAR00006, VAR00004, VAR00003.

^{b} For regression through the origin (the no-intercept model), [TeX:] $$R^{2}$$ measures the proportion of the variability in the dependent variable about the origin explained by regression. This CANNOT be compared to R2 for models which include an intercept.

^{c} Dependent variable: VAR00007.

^{d} Linear regression through the origin.

[TeX:] $$\text { Model summary }^{c, d}$$ | |||||
---|---|---|---|---|---|

Model | R | [TeX:] $$\mathbf{R} 2^{\mathrm{b}}$$ | [TeX:] $$\text { Adjusted } \mathbf{R}^{2}$$ | SE of the estimate | Durbin-Watson |

1 | [TeX:] $$1.000^{\mathrm{a}}$$ | 1.000 | 1.000 | 0.04035 | 1.855 |

Table 4 illustrates the importance of agricultural mechanization services and cost variance analysis at different wheat planting stages in Henan Province, as well as the mean square, degree of freedom, F-test, and significance level. The variance is 3797.878, the mean square 1265.959, the residual 0.122, and the associated probability Sig less than 0.05, indicating that the independent variable has a significant impact on the dependent variable. The importance of agricultural mechanization services at different wheat planting stages in Henan Province significantly impacts the cost.

Table 1.

[TeX:] $$\text { ANOVA }^{c, d}$$ | ||||||
---|---|---|---|---|---|---|

Model | Sum of squares | df | Mean square | F | Sig. | |

1 | Regression | 3797.878 | 3 | 1265.959 | 777559.213 | [TeX:] $$0.000^{a}$$ |

Residual | 0.122 | 75 | 0.002 | |||

Total | [TeX:] $$3798.000^{b}$$ | 78 |

^{a} Predictors: VAR00006, VAR00004, VAR00003.

^{b} This total sum of squares is not corrected for the constant because the constant is zero for regression through the origin.

^{c} Dependent variable: VAR00007.

^{d} Linear regression through the origin.

Table 5 shows that in the collinearity test, the t value of the cultivation stage V1 and the seeding stage V2 is 1.744E9, and the associated probability is 0. Thus, there is a severe collinearity problem between the two independent variables, and the dependent variables V1 and V2 need to be excluded from the model.

Table 5.

[TeX:] $$\text { Excluded variables }^{b, c}$$ | ||||||
---|---|---|---|---|---|---|

Model | Beta In | t | Sig. | Partial correlation | Collinearity statistics tolerance | |

1 | V1 | [TeX:] $$1.000^{\mathrm{a}}$$ | 1.744E9 | 0.000 | 1.000 | 3.215E-5 |

V2 | [TeX:] $$1.000^{\mathrm{a}}$$ | 1.744E9 | 0.000 | 1.000 | 3.215E-5 |

^{a} Predictors in the model: VAR00006, VAR00004, VAR00003.

^{b} Dependent variable: VAR00007.

^{c} Linear regression through the origin.

Table 6.

[TeX:] $$\text { Coefficients }^{\mathrm{a}, \mathrm{b}}$$ | ||||||
---|---|---|---|---|---|---|

Model | Standardized coefficients | Standardized coefficients | t | Sig. | ||

B | SE | Beta | ||||

1 | V3 | 0.302 | 0.013 | 0.293 | 22.807 | .000 |

V4 | -0.273 | 0.012 | -0.265 | -22.411 | .000 | |

V6 | 0.970 | 0.005 | 0.972 | 193.290 | .000 |

^{a} Dependent variable: VAR00007.

^{b} Linear regression through the origin.

Table 6 shows that the t values of the multiple linear regression equations for the plant protection stage V3, irrigation stage V4, and harvest stage V6 of Henan wheat are 22.807, -22.411, and 193.290, respec¬tively. Additionally, the associated probability is 0, less than 0.05, indicating linearity. After the regression equation is tested, the following regression equation models are obtained.

Eqs. (1) and (2) show that after the standardized coefficient, the mechanization of harvest stage V6 has the greatest impact on the cost of planting wheat, with a coefficient of 0.972. It has a significant positive impact on reducing planting costs, followed by plant protection stage V3 with a coefficient of 0.293, and then irrigation and drainage. The mechanization of stage V4 impacts planting costs, with a coefficient of -0.265, indicating that mechanized services in the irrigation and drainage stage may increase planting costs. But in general, the mechanized services in these three stages impact the cost of wheat cultivation in Henan Province.

Regression model of non-standardized coefficients:

Standardized regression model:

The social network analysis model (SNA) is a common social network analysis paradigm, describing the distance between the various related factors or stages and the clan relationship. In this study, UCINET 6.0 is used to complete the visual processing of social network phase relevance, visual processing of branch blocks, and centrality analysis of each phase, phase block cluster and phase block density analyses.

We choose SNA software (UCINET 6.0) for data entry, then use NETDRAW for visualization pro¬cessing and obtain the following diagram.

Spring embedding displays a strong correlation among tillage, sowing, plant protection, irrigation and drainage, fertilization, and harvesting (Fig. 1). The correlation coefficients are 1.000, 1.000, 1.000, 1.000, and 1.000, respectively. The cost of wheat has a similar effect. On this basis, we further analyze the centrality measures, using OutDegree and InDegree to characterize, we obtain the following data.

From the degree measures and centrality measurement in Table 7, it can be seen that the centrality of drainage and irrigation stage V4, fertilization stage V5, harvesting stage V6, tillage stage V1, sowing stage V2, OutDegree, and InDegree are all 4.107. The agricultural mechanization service cost of these five stages has an equally significant impact on the cost of the entire planting stage, while that of the plant protection stage V3 has a negligible effect. The OutDegree and InDegree centrality points are 0.535.

Table 7.

Order | Stage | 1 | 2 | 3 | 4 |
---|---|---|---|---|---|

OutDegree | InDegree | NrmOutDeg | NrmInDeg | ||

4 | V4 | 4.107 | 4.107 | 82.140 | 82.140 |

5 | V5 | 4.107 | 4.107 | 82.140 | 82.140 |

6 | V6 | 4.107 | 4.107 | 82.140 | 82.140 |

1 | V1 | 4.107 | 4.107 | 82.140 | 82.140 |

2 | V2 | 4.107 | 4.107 | 82.140 | 82.140 |

3 | V3 | 0.535 | 0.535 | 10.700 | 10.700 |

Table 8 reveals that wheat agricultural mechanization service cost is the relationship connection diagram and matrix. It can be divided into four blocks. The first block element is the cultivation stage V1, sowing stage V2, harvest stage V6, and drainage and irrigation stage V4. The second block element is the plant protection stage V3 and the fertilization stage V5. The interaction relationship between the first and the second block elements comprise the third and fourth blocks. The maximum value in the block is one, and the minimum 0.107.

Table 8.

Order | Stage | 1 | 2 | 6 | 4 | 5 | 3 |
---|---|---|---|---|---|---|---|

1 | V1 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.107 |

2 | V2 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.107 |

6 | V6 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.107 |

4 | V4 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.107 |

5 | V5 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.107 |

3 | V3 | 0.107 | 0.107 | 0.107 | 0.107 | 0.107 | 1.000 |

Table 9 shows that the maximum value is 1 among the four blocks. The density coefficient between the first and the second blocks indicates that the relationship between the two blocks is strong. The minimum value of the density coefficient between the first and the second blocks is 0.107, which weakens the relationship between the two.

Radial basis function (RBF) neural network is also called the local receptive field neural network. The principle is to use the RBF to perform interpolation operations in high-dimensional space, the RBF network model, and the backpropagation (BP) neural network. Both are multi-layered feed-forward BP networks. However, the RBF neural network is more accurate than the BP neural network. Therefore, it can overcome the slow process of BP network global approximation and accommodate more time-efficient models. RBF can detect the nonlinear and complex relationship between the input and the output layers to intuitively understand the six stages: cultivation stage V1, sowing stage V2, plant protection stage V3, drainage, and irrigation stage V4, fertilization stage V5, harvest stage V6, affecting the “cost of wheat in Henan.” We selected 78 data points (shown in Table 10) from Chongqing, Shaanxi, Henan, Liaoning, and other provinces and cities for model analysis on the complex influence of the relationship between the fertilization stage V5 and the harvest stage V6 on the total cost V7.

Table 10 shows that a total of 78 samples were selected, 5 of which were eliminated during the cal¬culation process. The number of model training samples was 55, the test support samples 18, and the effective rate 100%. Training is terminated when the error cannot be further reduced. Thus, the training effect is good and meets the standard requirements.

Table 10.

Case processing summary | N (%) |
---|---|

Sample | |

Training | 55 (75.3) |

Testing | 18 (24.7) |

Valid | 73 (100) |

Excluded | 5 |

Total | 78 |

The nonlinear and complex relationship diagram of the RBF network of the six planting stages of the “cost of wheat in Henan” (Fig. 2) shows that the model contains a hidden layer (Hidden-Layer) of four neural units (Units), an input layer (Input-Layer), and an output layer (Output-Layer). There are several lines mapped from the input layer to the hidden layer. The gray lines represent the positive weight relationships, and the blue line the negative weight relationships. Similarly, the mapping from the hidden layer to the output layer is the same. Again, the blue line represents a negative weight relationship, and the gray lines a positive weight relationship.

Fig. 3 shows that the independent variable is sorted by V1 at the tillage stage, with a coefficient of 0.222. The importance after standardization treatment is 100%; V2 at the sowing stage, with a coefficient of 0.199, standardized. The importance after treatment is 89.8%; V4 in the irrigation stage, the coefficient is 0.178, and the importance after standardization treatment is 80.2%; V6 in the harvest stage, the coefficient is 0.163, and the importance after standardization treatment is 73.4.%; V3 in the plant protection stage, the coefficient is 0.131, and the importance after standardized treatment is 59.1%; V5 in the fertilization stage, the coefficient is 0.106, and the importance after standardization is 47.5%.

As the characteristics of each research method are different, the calculation results of may be different as well. Therefore, we further analyze the calculation results of the three methods, multiple linear regression, SNA and RBF neural network, as shown in Table 11.

Table 11.

Importance | Multiple regression | SNA | RBF neural network |
---|---|---|---|

1 | cultivation stage V1 | cultivation stage V1 | cultivation stage V1 |

2 | sowing stage V2 | sowing stage V2 | sowing stage V2 |

3 | harvest stage V6 | harvest stage V6 | irrigation stage V4 |

4 | protection stage V3 | irrigation stage V4 | harvest stage V6 |

5 | irrigation stage V4 | fertilization stage V5 | protection stage V3 |

6 | - | protection stage V3 | fertilization stage V5 |

Table 11 shows the phased impact of agricultural mechanization services on wheat planting costs. From the perspective of multiple linear regression analysis, the dependent and independent variables of the cultivation stage V1 and the sowing stage V2 are completely collinear. The correlation coefficient between them indicates that the two independent variables have the most significant impact on the dependent variable. The most important result of the SNA is in the cultivation stage V1, followed by the sowing stage V2 and the harvest stage V6. The three stages are consistent with the multiple linear regression analysis results; the most important calculation result of the RBF neural network is also in the tillage stage V1, followed by the sowing stage V2.

However, overall, the first few results calculated by the three methods are the same: in the cultivation stage V1, the sowing stage V2, and the harvest stage V6. This indicates that the agricultural mecha¬nization services have the most significant impact on the cost of wheat cultivation. This is the experi¬mental result confirmed by research methods, original data, and test data.

Based on previous studies, this article analyzes the effect of agricultural mechanization on the cost of wheat in different planting stages by decomposing the wheat planting stages in Henan. First, we obtained the agricultural mechanization service cost as the dependent variable using the multiple linear regression model. Then, we obtained the linear equation with the cost of different planting stages as the independent variable to analyze the importance of mechanization services at different stages. Second, SNA was used to analyze the relationship between the impact of agricultural mechanization services and the cost of wheat planting at different stages, the results of visualization processing, centralization analysis, and density coefficient model analysis. Third, we analyzed the six planting stages of the “cost of wheat” through the RBF neural network model. We obtained the weight chart of the nonlinear complex mapping relationship among the input, the hidden, and the output layers. Finally, we compared and analyzed the factor importance ranking results of the three models of multiple linear regression, SNA, and RBF neural network, indicating that different research methods have different calculation results. Nonetheless, their factor ranking calculation results have convergence.

He completed his postdoctoral training at the School of Finance at the Shanghai University of Finance and Economics and received his doctoral degree at Tongji University’s School of Economics and Management. Currently, he is a professor of Finance at the Department of Economics and Finance, Yangtze Normal University.

- 1 H. Huang, Z. Yun, L. Zhang, Z. Wu, C. Ye, H. He, X. Chen, "Study on the characteristics of agricultural machinery purchase and leasing and management decision methods,"
*Chinese Journal of Agricultural Machinery Chemistry*, vol. 40, no. 5, pp. 193-200, 2019.custom:[[[-]]] - 2 J. Cai, W. Liu, "Analysis of the reasons for the development of agricultural mechanization and its service outsourcing: evidence from questionnaire survey data in the three provinces of Hebei, Henan, and Shandong,"
*Chinese Journal of Agricultural Resources and Regional Planning*, vol. 39, no. 2, pp. 230-236, 2018.custom:[[[-]]] - 3 J. Pan, P. Li, C. Chen, Q. Meng, "Impact of agricultural mechanization services on corn production efficiency: an empirical analysis based on panel data of main corn production areas from 2004 to 2017,"
*Journal of Chinese Agricultural Mechanization*, vol. 41, no. 6, pp. 210-215, 2020.custom:[[[-]]] - 4 S. Y u, T. Liu, B. Cao, "The impact of agricultural mechanization services on food production cost efficiency: empirical evidence from the main wheat producing regions in China,"
*Journal of Huazhong Agricultural University (Social Sciences Edition)*, vol. 2019, no. 4, pp. 81-89, 2019.custom:[[[-]]] - 5 Z. Zhao, J. Ren, S. Li, Y. Wang, "Can agricultural machinery services improve the technical efficiency of wheat production? Based on an empirical analysis of the 2007-2017 provincial panel data,"
*Journal of China Agricultural University*, vol. 25, no. 11, pp. 150-161, 2020.custom:[[[-]]] - 6 C. Chen, Y. Li, X. Liao, "The productivity effect analysis of rice production outsourcing: based on panel data of three counties in Jiangsu Province,"
*Chinese Rural Economy*, vol. 2012, no. 2, pp. 86-96, 2012.custom:[[[-]]] - 7 Z. Zhang, Z. Yi, "Research on the impact of agricultural productive service outsourcing on rice productivity: based on the empirical analysis of 358 farmers,"
*Issues in Agricultural Economy*, vol. 36, no. 10, pp. 69-76, 2015.custom:[[[-]]] - 8 H. Zhou, Q. Wang, Q. Zhang, "The aging of rural labor force and the lack of rice production efficiency: based on the perspective of socialized services,"
*Chinese Journal of Population Science*, vol. 2014, no. 3, pp. 53-65, 2014.custom:[[[-]]] - 9 Y. Ji, F. Zhong, "Non-agricultural employment and the utilization of farmer’s agricultural machinery service,"
*Journal of Nanjing Agricultural University (Social Sciences Edition)*, vol. 13, no. 5, pp. 47-52, 2013.custom:[[[-]]] - 10 H. Song, C. Jiang, "Research on farmers’ choice of agricultural machinery socialized services: based on a questionnaire survey of wheat growers in 8 provinces,"
*Journal of Agrotechnical Economics*, vol. 2015, no. 9, pp. 27-36, 2015.custom:[[[-]]]