1 Introduction

According to characteristics of tea industry information service, and based on analysis of the evaluation goal for tea industry information service quality, the typicality of index system as the main direction, also takes the information requirements characteristic of users into account, the key factors influencing the quality of information service have been analyzed from platform function, platform information, platform design, platform framework and platform efficiency, then select 5 first grades and 30 s grades. 30 s grades are clustered by R-cluster analysis, finally forming an evaluation index system consists of 5 first grades and 17 s grades. Based on the influence factor of information service quality analyzed by information service quality evaluation model built with multiple regression, which will give a scientific guidance for tea industry information service.

2 Analysis Methods

2.1 R-cluster

R-cluster [1] is one kinds of the cluster analysis method, is adequate for lots of variable, and variable with a high correlation, its purpose is to cluster the variables in the same group, and find the representative variables to achieve the purpose of dimension reduction. In R-cluster theory, N objects are treated as N groups, the correlation coefficient around the groups is R, two groups having the biggest related coefficient are clustered in one groups, until all groups are clustered in one big group.

2.2 Multiple Linear Regression

2.2.1 Brief Introduction for Multiple Linear Regression Model

Multiple linear regression [2] model is one kinds of math model to consider the relationship around variable x1, x2,…, xp and variable y. Assuming y is change along with the changes of x1, x2,…, xp.

The mathematical expression of multiple linear regression is Y = W(3) * X + C. W(3) is weight matrix, C is constant term.

2.2.2 Selection Criteria of Independent Variable for Multiple Linear Regression Model

Multiple linear regression model has advantages, such as simple method and high precision, when it simulate the data owning pure linear relationship, but having bad results in this situation of data is nonlinear relationship or data missing. In order to guarantee the excellent predictive results, the selection criteria of independent variable are summarized as follows: (1) independent variable has a outstanding influence with dependent variable, and there is a close linear relationship; (2) linear relationship between independent variable and dependent variable is true, not in form; (3) a certain mutual exclusion between the independent variables is necessary; (4) independent variable must have integrated data [3].

3 Index System

3.1 Preliminary Index System

Selecting platform function, platform information, platform design, platform framework and platform efficiency as the first grade indexes [4], and every index is further segmented, finally forming 30 s grade indexes, which forms the service quality index system for tea industry information platform, the specific indexes are listed as follows:

3.2 Index System Selection

Normally, the method of principal component analysis is used to solve a problem which contains lots of evaluation indexes, but the results from principal component analysis depend on measured value of primitive index, so principal component analysis only solve simplification of quantitative index, not the descriptive index before quantitative [5].

In this paper, R-cluster is adopted. According to the correlation coefficient in indexes, all indexes are clustered by R-cluster, the demand of accuracy finally determines how many groups will be segmented, select the typical index in one group represent multiple indexes to achieve the purpose of simplified index [6].

Assuming there are P indexes u1, u2, u3, …, up; N experts score for the relative coefficient among the indexes, the relative coefficient matrix after scoring is expressed in Table 2, and the cluster results are expressed in.

To combine the results of cluster, comprehensively considering the simplicity, representation and feasibility of indexes, analysis results are carried out as follows: when the minimum distance of correlation coefficients is under 0.03, the jumping points are obvious, indexes are clustered 17 groups. So, the results represented as follows: R1: 4, 11, 15, 7, 13, 20, 8, 18, 21; R2: 3, 23; R3: 5, 17; R4: 6, 10, 14, 30; R 5: 1; R6: 2: R 7: 9; R8: 12; R9: 24; Rl0: 28; R11: 19; Rl2: 27; R13: 16; R14: 21; R15: 22; R16: 26; R17: 29 (Fig. 1).

Fig. 1.
figure 1

Result of R-mode clustering

Full size image

4 Data Acquisition

4.1 Experts Estimation

After the establishment of indexes system, expert estimation has two parts:

  1. (1)

    General part indexes existing in traditional information service website or system. After the establishment of indexes system, select the typical tea information website, then let the experts who is good at agricultural informationization estimate the quality of information service websites and the indexes that influencing the quality of information service.

  2. (2)

    The indexes part reflecting the characteristics of cloud service platform. Because of tea industry information cloud service platform in domestic is still blank, there is no quantitative index to be verified, only the experts’ view as the reference, that is means, when tea industry information cloud service platform is built, the 7 indexes, such as X23: Platform stabilization, X24: Platform compatibility, X25: Platform safety, X26: Platform adaptability, X27: Platform expansibility, X28: Software deployment simplicity and X29: Software operation shortcut will be fully considered.

4.2 Data Processing

First step: website selection. Tea information websites are divided into two groups: one group is comprehensive agriculture websites but covering content of tea industry information, another group is the professional tea information websites. The objects of this experiment are 40 websites in two groups, 18 comprehensive websites, such as agriculture information websites belong to province and municipal district, and 22 professional tea information websites, specific situation is listed in Table 3.

Second step: data acquisition. Experts estimate the 10 s indexes belong to Platform function, Platform information and Platform design in Table 1, information service quality and indexes estimated by quantitative evaluation, the scores are during 0–1, adopt the method of integrating multiple scores, the data is listed as Table 4 after data processing.

Table 1. Service quality evaluation index system of information platform for tea industry
Full size table
Table 2. Related coefficient matrix
Full size table
Table 3. Objects of experimental evaluation for tea information sites
Full size table
Table 4. Experimental data
Full size table

5 Evaluation Model

5.1 Experiment Data Processing

In experiment, 30 data before in Table 4 as the training sample, 11 data after in Table 4 as the test sample.

5.2 Selection of Influencing Factors

Before the influencing factors are finally selected, the correlation between information service quality and influencing factors needed to be analyzed respectively. The purpose of correlation analysis is to represent the correlation situation and its change rule between variables, and find the correlation model with each other, which can provide references for making the next decision.

Null hypothesis of test is correlation coefficient of two variables expressing 0 in totality. The process of correlation analysis in SPSS shows the probability of hypothesis formation [7].

On the SPSS 18 platform, data correlation analysis respectively is produced between Y and X1, X2, X4, X5, X9, X10, X12, X16, X19, X22. Correlation between Y and X2, X4, X5, X9, X10, X16, X19, respectively is obvious, and shows linear relationship, and these indexes can be selected as the influencing factors for Y, so multiple regression model is established.

5.3 Experiment Analysis

On the platform SPSS 18, multiple regression model has been established, X2, X4, X5, X9, X10, X16 and X19 as independent variables, Y as dependent variable. Results of multiple linear regression show that X2, X4, X5, X9, X10, X16, with Y shows the positive correlation respectively, but the estimated value of regression coefficient for X19 shows negative correlation, which is not fact, so estimated value of X19 is not believable [8], then eliminate variable X19. Second regression is carried out.

R2 of model is 0.999, adjusted R2 also is 0.999. Value of F tends to rationality. To a certain extent, model reflect the relationship for information service quality and influencing factors, finally the influencing factors model of information service quality is carried out as follows:

$$ {\text{Y}} = 0.0 7 5+ 0. 100*{\text{X}}_{ 2} + 0.00 5*{\text{X}}_{ 4} + 0. 3 6 8*{\text{X}}_{ 5} + 0.0 2 5*{\text{X}}_{ 9} + 0. 4 8 1*{\text{X}}_{ 10} + 0.0 20*{\text{X}}_{ 1 6} . $$

5.4 Model Test

The established multiple regression model is tested by test data.

Table 5 shows that results of 10 test data present ideal results, relative errors between predictive value and real value all are less than 3 %, and average error is only 1.01 %. Experiment demonstrates that the established model is multiple fitted equation having a good reliability [9].

Table 5. Relative errors between predicted and actual values
Full size table

6 Conclusions

All study above shows that two aspects will be paid more attention to during the building of cloud service platform:

  1. (1)

    In the process of construction for information service system, these indexes such as service interactivity, service individuation, service professionalism, information vividness, information comprehensiveness and interface friendly, are important to influence information service quality.

  2. (2)

    During the process of building for cloud service platform, these 7 indexes such as platform stabilization, platform compatibility, platform safety, platform adaptability, platform expansibility, software deployment simplicity and software operation shortcut should be fully considered.