Dataset for Transformer-Based Time Series Forecasting of Credit Limit Utilization and Risk Profiling in Retail Lending

This dataset supports the findings of the article “Transformer-Based Time Series Forecasting for Credit Limit Utilization and Risk Profiling in Retail Lending.” The dataset contains anonymized and preprocessed time series data used to develop and evaluate transformer-based forecasting models. The data were prepared exclusively for research purposes. All personally identifiable information has been removed or anonymized in compliance with applicable data protection and privacy regulations. The dataset is intended for academic and non-commercial research in time series forecasting and financial risk modeling. In addition to the core input data, the repository includes: - a forecasted balance time series file, which was used to construct the visualizations presented in Figure 3 and to compute the evaluation results reported in Table 2; - a risk rating binning file These supplementary files are provided to enhance transparency, reproducibility, and interpretability of the experimental results reported in the article.