3.1. Basic Model of DEA
The DEA method requires finding an optimal set of input–output weights. The basic model of the DEA method is as follows:
and represent unknown output and input weights; represents a specific bidder in bidders (); and each bidder has input metrics () and output indicators ().
The DEA method aims to determine the optimal output and input weights for the owner to achieve the desired target. This target represents the optimal benefit between each input and each output of all bidders (for example, the smallest input gets the largest output). This problem can be converted into a linear programming problem:
Through solving the above linear programming problem, the optimal solution of
and
(call it
and
) can be obtained. Therefore, the efficiency value of each bidder, named self-evaluation value, can be calculated as follows:
Using (3), the efficiency value of each bidder can be calculated; (1) indicates the bidder is valid, and (2) indicates the bidder is invalid. Under normal conditions, multiple bidders exist whose reaches 1. Therefore, constraints must be added to further evaluate the bidders.
This research will further analyze several methods which improve the relative optimal efficiency DEA evaluation model, including cross-evaluation mechanism, balance index, and comprehensive input efficiency.
3.2. Cross-Evaluation Mechanism
The basic idea of cross-evaluation mechanism is as follows: the optimal solutions
and
of each bidder are used to calculate the efficiency value of other bidders, obtaining cross-evaluation values:
For the above issues, the higher the cross-evaluation value
, the more favorable it is for bidder
, and the more unfavorable for bidder
. The above questions can be converted into
Through solving the above linear programming problem, a new set of optimal solutions for
and
(call them
and
) can be obtained. The cross-evaluation value and cross-evaluation matrix are obtained by using the optimal solution:
The relative efficiency value
of the bidder obtained above is the value on the diagonal in the cross-evaluation matrix
, which serves as the first evaluation index. The second evaluation index is divided into two kinds. Many researchers use the average value
of off-diagonal elements in line
as the second index of evaluation. Additionally, the smaller the
, the better. Additionally, much research uses the average value
of the element in column
as the second index of evaluation. Additionally, the bigger the
, the better. In the evaluation process of DEA, the first evaluation index
is compared first. If
is the same, then the second evaluation index
is compared. This research takes the average value
of non-diagonal elements in column
as the second index of evaluation. The calculation formula of
and
are as follows:
In the process of bid evaluation, the efficiency of bidders is compared first. Additionally, bigger is prioritized. If is identical, or is compared, with bigger or smaller being given priority. For the second evaluation index , it takes the average value of off-diagonal elements in line as the evaluation index. However, in the calculation process of DEA evaluation method, all calculations are carried out for each column of data, aiming to find the optimal weight distribution between input and output of each column. If is used as the second evaluation index, reasonable results cannot be obtained in most cases. Thus, it is not suitable to use as the second evaluation index. Thus, , which takes the average value of non-diagonal elements in column, should be the second evaluation index.
When
is the evaluation index, in most cases, relatively complete evaluation results can be obtained. However, in special situations, the same value of
may also occur, which cannot offer complete evaluation results. As a result,
should be chosen as the second evaluation index. In view of the possibility that
may have the same value, another cross-evaluation should be conducted. The new optimal solutions
and
should be used to calculate the efficiency value of other bidding units to obtain a new cross-evaluation matrix. The second cross-evaluation is referred that the average value of non-diagonal elements in column
is used as the evaluation index (the evaluation index of the second cross-evaluation is expressed by
). The calculation process of DEA model with cross-evaluation mechanism is shown in
Figure 1.
The complete evaluation can be completed after a cross-evaluation process or after a second cross-evaluation process if the number of bidders i is relatively small. However, if there are a large number of bidders, a second cross-evaluation process may not be able to provide a complete judgment. Therefore, to obtain a comprehensive evaluation report, multiple cross-evaluation methods must be employed.
Based on the above analysis, it appears that the cross-evaluation mechanism can assist in evaluating and ranking bidders (DMUs) more accurately. It is important to note, however, that for large numbers of bidders, DEA evaluation with the cross-evaluation mechanism may require multiple cross-evaluations to produce reasonable results.
3.3. Balance Index Model
For the above DEA bid evaluation model, it can be known that
is the total output of the
bidder and
is the total input of the
bidder. Then, the interest constraint conditions of the
bidder are as follows.
Assuming that
is an efficiency function, then
In addition, under the condition of maximizing competitive interests, the input of the bidder is proportional to , and the output of the bidder is proportional to . That the input of the bidder is zero is called the ideal optimum.
In the process of bid evaluation, when the interest constraint condition of the
bidder is zero, it is said to be valid; otherwise, it is said to be invalid. If the interest constraints of other bidders are not equal to zero, then the
bidder wins. If the interest constraints of several bidders are equal to zero, it is necessary to introduce “balance index” [
61] for further evaluation and analysis. Balance index of each bidder refers to the sum of all other bidders’ interest constraints. In the process of bid evaluation, the smaller the balance index is, the better. The balance index is expressed by
:
where
is the optimal input weight of the
bidder;
is the input of bidder
. The indicators for the other parameters are as described previously.
In DEA evaluation model, which introduces balance index, the efficiency value of each bidder is compared first. Additionally, the higher the , the higher the priority. If is identical, then the balance index is compared. Additionally, the smaller is, the higher priority will be given. In this method, complete evaluation results can be well given even if the number of bidders is relatively large. The advantage of this method is that it is only necessary to calculate the balance index of the bidder (DMU) once to obtain a reasonable evaluation result.
3.4. Comprehensive Input Efficiency Model
By referring to the principle of cross-evaluation mechanism, the optimal input weight
of each bidder is used to calculate the cross-evaluation value and cross-evaluation matrix of input efficiency of other bidders:
The efficiency value
of the bidder was obtained previously. In Equation (13), all the diagonal elements in the cross-evaluation matrix
are 1. The efficiency value of each bidder is the minimum value of all elements in column
, denoted by
, which is equal to the efficiency value calculated above:
If only self-evaluation value
is used for evaluation, the same situation may occur in
, which will lead to failure to evaluate all bidders completely. Thus, this study presents the concept of comprehensive input efficiency, which is expressed by
:
Through the DEA evaluation model, which introduces comprehensive input efficiency, in most cases, the bidder can be completely evaluated. The priority order of each bidder can be obtained by comparing the size of directly. However, in special cases, and can be combined to evaluate bidders. First, is compared. If has the same value, then is compared. Additionally, the bigger , the more priority it has. Comprehensive input efficiency requires only one linear programming calculation to rank the efficacy of all bidders. Multiple cross-evaluations are not required. Moreover, this model is more efficient in its evaluation and will not be affected by the number of bidders .
The calculation process of DEA model that introduces difference index or comprehensive input efficiency is shown in
Figure 2.
In this section, the DEA basic model is introduced first. Then, cross-evaluation model and balance index model were analyzed. Finally, based on the above analysis, a new DEA ranking model, the comprehensive input efficiency model, is proposed. As a result of the analysis, it is evident that in the cross-evaluation model, multiple evaluation processes are required to achieve a complete ranking. However, the balance index model and the comprehensive input efficiency model only need to calculate one linear programming equation and the second evaluation index or to obtain more complete ranking results. Thus, the bid evaluation efficiency of the latter two models is higher.