Multi-criteria decision-making (MCDM) is a process used to evaluate and rank different alternatives based on several criteria, which may have different levels of importance and may conflict with one another.

### Abstract
The abstract presents GPU-TOPSIS, a parallel and vectorized implementation of the TOPSIS multi-criteria decision-making method designed for very large datasets. The proposed method uses GPU tensor operations to accelerate normalization, weighting, ideal-solution calculation, distance computation and ranking. It also introduces a two-pass fragmentation strategy that allows datasets larger than the available memory to be processed while preserving the mathematical results of traditional TOPSIS. Experiments using Amazon product data show speed improvements of up to 4.75 times compared with a CPU implementation, while maintaining nearly identical scores and rankings across CuPy, PyTorch and TensorFlow.