Performance Benchmarking of Non-banking Financial Institutions by Means of Self-organising Map Algorithm

We construct a benchmarking model in the form of a twodimensional self-organising map (SOM) to compare the performance of nonbanking financial institutions (NFIs) in Romania. The NFIs are characterized by a number of performance dimensions such as capital adequacy, assets’ quality and profitability. First, we apply Kohonen’ SOM algorithm (an unsupervised neural network algorithm) to group the NFIs in clusters with similar characteristics. Then, we use the U-matrix method to build maps that facilitate the visualization of SOM results and select the best map in terms of quantisation error and ease of readability. The best map is used to analyze the companies over time by studying the cluster where each company was positioned for each period. We conclude that there are benefits in using SOM for interpreting large and complex financial data by identifying and visualizing clusters.