Decision Making Method for Non-Independent Hierarchical Structures (NP Method)

MATLAB 4K 244 views 0 downloads 1 credits
Tags:

Resource Overview

The NP (Network Process) method, developed by Professor T.L. Saaty at the University of Pittsburgh in 1996, is an adaptation of the Analytic Hierarchy Process (AHP) specifically designed for non-independent hierarchical decision structures. This practical decision-making methodology involves key computational steps such as: 1) Factor identification and model establishment for network and control layers, 2) Creation of comparison matrices, 3) Column-wise normalization based on indicator types, 4) Calculation of maximum eigenvalues and corresponding eigenvectors for each comparison matrix, 5) Consistency validation with matrix element adjustment when required, 6) Unitization (normalization) of eigenvectors to form judgment matrices, 7) Multiplication of control layer and network layer judgment matrices to obtain weighted supermatrices, 8) Limit matrix computation through power iteration convergence.

Detailed Documentation

The NP method, proposed by Professor T.L. Saaty from the University of Pittsburgh in 1996, is a decision-making approach specifically designed for non-independent hierarchical structures. Compared to the traditional Analytic Hierarchy Process (AHP), this new practical methodology demonstrates enhanced decision analysis capabilities. Below are the detailed implementation steps:

1. Identify factors and establish models for both network and control layers. This fundamental step is crucial as it determines the method's uniqueness and practicality. In code implementation, this typically involves creating hierarchical data structures with dependency mappings.

2. Create comparison matrices by evaluating each factor against all others and assigning scores based on comparative importance. Algorithmically, this can be implemented using pairwise comparison functions that generate n×n matrices where n represents the number of factors.

3. Perform column-wise normalization by dividing each column element by the column sum. This preprocessing step, often implemented using vector division operations, prepares the matrix for subsequent eigenvalue calculations.

4. Calculate the maximum eigenvalue and corresponding eigenvector for each comparison matrix. This computational step, typically solved using numerical methods like power iteration, determines each factor's relative importance within the hierarchical structure.

5. Conduct consistency validation. If the comparison matrix fails consistency conditions, adjust relevant elements accordingly to ensure rational decision outcomes. Implementation often involves calculating consistency ratios (CR) and employing optimization algorithms for matrix refinement.

6. Normalize individual eigenvectors to unit length and compile them into judgment matrices. This standardization process, accomplished through vector scaling operations, facilitates subsequent matrix computations.

7. Multiply the control layer's judgment matrix with the network layer's judgment matrix to obtain a weighted supermatrix. This matrix multiplication operation determines the comprehensive weights of factors throughout the hierarchical structure.

8. Normalize the weighted supermatrix and compute its limit matrix through K-th power convergence. The j-th column of the resulting matrix represents the limiting priority vector of network layer elements relative to element j. This final computation, often implemented using iterative matrix multiplication until convergence, establishes the ultimate ranking of factors for final decision-making.

In summary, the NP decision method represents an efficient and practical approach that significantly enhances decision-making processes, particularly demonstrating extensive application value in non-independent hierarchical structure decision problems.

You May Also Like