Candidate of technical sciences Р. V. Tereliansky

Volgograd state technical university, Russia

Distributed computer decision support system

The distributed computer decision support system was developed on the basis of the method of the analytic hierarchy process. Method makes it possible to divide many experts, who participate in the examination, by several subsets depending on subject area and nature of the criteria, using in the hierarchy. Geometric mean (1) (where – the aggregated estimation of the element, which belongs to i line and j to the column of the matrix of paired comparisons; n – number of matrices of the paired comparisons, each of which is by one expert) is using for aggregation of the opinions of experts. The averaging of the judgments of experts is realized according to their qualification in the sufficiently critical tasks with the justified expenditures for examination. The calculation of the aggregated estimation in the case of the recruiting n experts, who have different importance, is achieved according to formula (2), where – the estimation of object, carried out by k-expert with the weight coefficient ; in this case . The methods of the theory of grade correlation are used during treatment of the data of collective expert estimation. The coefficient of concordance V makes it possible to estimate the degree of the preferability series’ coordination between themselves, which are built by each expert. The value of the V-coefficient must be within the limits . V=0 indicates complete opposition, V=1 – the complete agreement of rankings. Practically authenticity is counted as normal, if V=0.7,...0.8. The small value of the coefficient of concordance, which testifies about the weak coordination of the opinions of experts, is the consequence of the following reasons: there is no generality of opinions in the totality of experts, actually; there are groups with the high coordination of opinions inside the totality of experts, but the generalized opinions of such groups are opposite. For the clarity of idea about the degree of the coordination of the opinions of any two experts А and В there is the coefficient of pair-rank correlation. The coefficient of pair-rank correlation takes the values . Value corresponds to the complete agreement of estimations in the ranks of two experts (complete coordination of the opinions of two experts). Value corresponds to two opposite rankings of the importance of the properties (the opinion of one expert is opposite to the opinion of another). The distributed computer decision support system making on the basis of Internet-technology is divided into two parts: client and server. Client part loads from the Internet. Experts are usually working with this part of the system (they are filled the matrices of paired comparisons). Server part assumes and processed data, obtained from the expert. The server part of the program system determines the global vector of priorities on the basis of this data, evaluates the degree of rank correlation, achieves averaging of the opinions of several experts and contains the block, which corresponds for the distribution to the experts of information’s about the degree of the coordination of their judgments. Figure 1 depicts the block diagram of client part. The description of each block is given below.

1. The expert receives an e-mail with the proposal to participate in the collective examination. In the letter it is contained: the identifier of expert (login), the code of access and www-address.

2. The description of the decided task is located on www-address.

3. To expert has to fulfill the matrices of paired comparisons.

4. The block of enumerating the uniformity will report this disturbance to expert if the uniformity of judgments is disrupted during the filling of matrices.

5. The e-mail with the results of examination is formed after successful producing of the estimation of problem. The e-mail contains: the identifier of expert, the identifier of problem, many identifiers of matrices and matrices of paired comparisons, eigenvectors of matrix and the indices of the relation of uniformity.

6. The block (which is corresponds for the dispatch of letters) sends data of server part after the composing of e-mail.

Fig.1. The block diagram of the client part of the program system

Server part is represented in figure 2 in the form the graph of the functioning of the computer system. The description of the units of the graph (number of unit in the figure it corresponds to the number of function in the brackets) is given below. Initialization of program system (1). Obtaining and the interpretation of e-mail (2). Fulfilling the data base (3). Load of the data about experts (4). Load of the matrix of paired comparisons (4). Determination of rank on the basis of the eigenvectors of each matrix of paired comparisons (5). The coefficient of concordance V is calculated (6). If V> 0.7, then goes to unit 12. The block of determining of the expert, whose opinion has the low degree of coordination with other experts (7). The block, which is making it possible to make the selection: to send away the e-mail to the expert (should go to unit 10) or removed uncoordinated data, obtained from the expert, which means, one should switch over to unit 11 (8). Sending of the e-mail to an expert with the recommendation to analyze problem one more time (9). The removal of data which were obtained from the expert, whose opinion is strongly differed from the opinions of remaining experts (11). Selection of the method of decision making: taking into account the degree of confidence to expert or without (12). Passage to block 13 is accomplished taking into account the degree of confidence in the expert. Otherwise, go to unit 14. Estimation of the degree of confidence to the opinion of experts (13). Selection of the conformable matrix of the paired comparisons (MPC), sent by experts (14). Calculation of the aggregated estimation of the elements of MPS according to formula 2 (15). Calculation according to the formula 1 (16). Location of the global vector of the priorities of decided problem (17). Conclusion of results and the completion of the work of application (18-19).

Fig. 2. Graph of the functioning of the server part of the program system