How ANP Overcomes AHP Limitations: A Step-by-Step Guide with Real-World Example

ANP (Analytic Network Process) removes the restrictive assumptions of AHP, allowing decision makers to consider complex dynamic interactions among elements.

Basic Structure of ANP

Supermatrix Algorithm

In an ANP network, control‑layer elements are denoted as the controlling set, while the network layer consists of element groups. The supermatrix is constructed with rows representing "receiving" elements and columns representing "source" elements. Pairwise comparisons are performed to obtain relative preferences and importance, yielding a normalized eigenvector whose columns sum to 1, although the matrix itself is not normalized.

The weighted supermatrix is obtained by multiplying matrix A with matrix W. To reflect dependencies, the weighted supermatrix undergoes a limit process, computing the limit relative ranking vector. If convergence is unique, the j‑th column of the limit matrix gives the final ranking of elements under the control element.

ANP Decision Procedure

Perform pairwise comparisons among all interacting elements in the network model.

Construct the final weighted supermatrix by inserting normalized eigenvector values into the supermatrix columns.

Normalize the columns of the supermatrix to ensure each sums to one.

Compute the weighted supermatrix.

Calculate the limit supermatrix using the power method (repeated multiplication until column vectors stabilize).

Case Study

The example evaluates the relationship among cost, maintenance, durability, and car categories. Weight matrices are derived, an initial supermatrix is built, and a sample weighted supermatrix A=[0.5,1;0.5,0] is used. After 4‑6 self‑multiplications with column normalization, a stable limit supermatrix is obtained.

The ANP results indicate that the American car is the optimal choice, with cost being the decisive factor.