The Application of Ahp in Electric Resource Evaluation

This article utilizes the analytic hierarchy process (AHP) method to study and establish the hierarchy model and its evaluation system for the electric resource evaluation.


Introduction
In recent years, with the continual increase of electric resource, the purchasing of electric resource has becoming the important part of the construction of library literature resource.The outlay used in electric resource in many libraries has exceeded 25%, so whether for purchasing new electric resource or continuing order and maintaining present electric resource, a new problem occurs, i.e. how to select and evaluate the electric resource, and how to enhance benefits and make the construction of electric resource more reasonable and scientific through the evaluation of electric resource?However, present scholars' researches only focused on the evaluation of internet information resource or single database, and there are few researches to study the integrated evaluation to the collection of electric resource, and the evaluation and system are not perfect, the present evaluation methods include qualitative method and quantitative method.In this article, we will adopt the AHP method which combines qualitative analysis with quantitative analysis, and try to establish the electric resource evaluation system that can be applied in the pre-evaluation before purchasing and the after-evaluation in the purchasing.
The AHP method (Cai, 2005, p.58-63) was put forward by US famous operational research expert T. L Saaty in 1970s, which tried to simulate three human basic characters (i.e.decomposing, estimation and integration) to deal with complex problems through analytic hierarchy, quantitative analysis and standardization, and added statistical test in the whole process.It adapts to solve those decision problems that have complex structure and many decision rules and are difficult to be quantified.
The basic approach of AHP method include following steps.(1) Establish the concept of the complex problem and find out main factors involved in the study objective.(2) Analyze the association and subjection relationships among factors and establish orderly ladder hierarchy model.(3) Compare both relative essentialities of various factors on the same layer to the certain rule on the upper layer, and establish the evaluation matrix.(4) Compute the relative weight of the compared factor to the rule on the upper layer according to the evaluation matrix and implement the coherence test.( 5) Compute the integrated weight of various layers to the total objective of the system and implement total compositor of the layers.

Constructing comparison evaluation matrix
We adopt the 1-9 standard degree method (seen in Table 1) to evaluate the relative essentialities of the indexes, and evaluate the proportion degree of the relative essentiality through both comparison among them.
For example, for electric resource, we think the content of database is comparatively more important than the searches system and function and endows it 3 points, and it is more important than the uses and endows it 4 points, and it is little more important than the cost accounting and endows it 2 points, and it is more important than the manufacturer than manufacturer service and endows it 4 points, and in this way, we can establish A-B evaluation matrix (seen in Table 2).

Computing the weights W I of various indexes
The information base of AHP is the evaluation matrix, and it utilizes the compositor principle to obtain the matrix compositor vector and compute the weight coefficients of various indexes.The computation approach (Li, 2004, p.75-78) includes following steps.
(1) Compute the product M i of factors on every raw of the evaluation matrix B: (2) Compute the root of n of M i on every raw: w i = n i M , i=1, 2… n, and n is the order number of the matrix in the equation.
(3) Implement normalized processing to (w l , w 2 ...wn) T , and make , …Wn] T are the eigenvectors, i.e. the weighted coefficients of various indexes.
The concrete evaluations of various layer index weights are seen in Table 4.

Implementing coherence testing to various evaluation matrix
Because of the complexity of things and human difference of objective evaluation, every evaluation can not achieve completely identical, and to ensure the rationality of the conclusion of AHP method, we need to implement coherence test to various evaluation matrix, so we introduce the negative square values of other latent roots except for the maximum latent root of the evaluation matrix in AHP method and take them as the deviation coherence index of the matrix departures, i.To test whether different evaluation matrixes have satisfactory coherence, we must introduce the average random coherence index RI value of the evaluation matrix, and RI values of 1-9 order evaluation matrixes can be seen in Table 3.
To the 1st and 2nd evaluation matrixes, they always have satisfactory coherence, but the order number exceeds 2, the ratio of the coherence index CI with the some order random coherence index RI is called the random coherence ratio CR, and when CR= RI CI < 0.10, we think this evaluation matrix has satisfactory coherence, i.e. the thinking on various layers is coherent, the conclusion obtained by the AHP is coherent, or else, the evaluation matrix should be adjusted to make it possess satisfactory coherence.
The maximum latent roots of the evaluation matrixes are obtained by the computer program, and according to them we can obtain the coherence index CI and random coherence ratio CR, and then we implement the coherence test, and the results are complete coherence or satisfactory coherence.

Integrated weight
Though above approach, we can only obtain the weighted vectors of a group of factor to the certain factor on its upper layer.To obtain the relative weights of various factors to the total objective, especially to obtain the compositor weights of various indexes on the lowest layer to the objective, i.e. "integrated weights", we need superincumbent computation and integrate weights under the single rule, and obtain the relative weight of every evaluation objective in the layer objective to the total objective and implement total evaluation coherence test layer by layer.Relative to the total objective, the integrated weights of various indexes can be denoted as W=a i a ij a ijk , where a i , a ij and a ijk respectively are 1st, 2nd and 3rd class index weight.Then we implement total compositor to the relative weights.
So we can obtain a clear evaluation index system of electric resource (seen in Table 4).
Though confirming the evaluation grading (seen Table 5), we can evaluate various indexes of the electric resource evaluation system, and the method is to use the weighted adding method, multiply the evaluation value of every evaluation index with the corresponding weight of this index, obtain the weighted evaluation value of the index, add these weighted evaluation values and obtain the total evaluation valves of the evaluation objective.The formula is S=∑W i P i (here, W i is the integrated weight of the i'th index, P i is the evaluation value of evaluated object on the i'th index, and i is the sequence number of the concrete index on the lowest layer in the evaluation model).
If the quantity of electric resources participated in the evaluation has Ql, Q2, Q3…Qn, we should adopt the AHP method.We respective establish evaluation matrixes aiming at 63 evaluation indexes, and obtain the compositor vectors of 63 evaluation matrixes.Multiply every compositor vector with weighted coefficient of corresponding index and add them, we can obtain the total compositor vector of Ql, Q2, Q3…Qn, and its result is also the compositor of the electric resources Ql, Q2, Q3…Qn.

Conclusions
The character of AHP is to combine qualitative analysis with quantitative analysis, which has high validity, reliability, conciseness and extensive applicability.But AHP still has limits, and its result only aims at the evaluation index in the rule layer, so the confirmation of evaluation index largely influences the system evaluation, in addition, human subjective evaluation has certain influence to the evaluation results of the system, so this method usually is combined with Delphi method to confirm the values of various indexes.
The evaluation of electric resources by AHP compensates human limit of subjective blur ability, which quantifies decision-makers' experiences and judgments, compares relative factors layer by layer, tests the rationality of comparison result layer by layer, avoids the subjective random of simple evaluation to make the result more exact and make the evaluation decision possess more objectivities and persuasions.
coherence of the evaluation thinking.The maximum latent root max

Table 2 .
Evaluation matrix

Table 3 .
RI values of evaluation matrix The numbers in the bracket are the weights relative to superior indexes.