A study of Data-Driven Forecasting of PISA Reading Literacy Performance of Türkiye: Artificial Intelligence Based Time Series Estimation. Yıldırım et al.

The study compares GRU, LSTM, ARIMA, and Prophet models using recursive multi-step forecasting and sequential variable inclusion strategies. The repository includes raw and processed datasets, missing data handling procedures, evaluation metric outputs (MAE, MSE, RMSE, DTW), forecasting results for the 2025, 2028, and 2031 PISA cycles, reproducible Python scripts, and supplementary methodological materials.

Independent variables used in the forecasting framework include parental education (PARED), cultural possessions (CULTPOS), highest parental occupational status (HISEI), and sense of belonging at school (BELONG). Missing values were completed using interpolation and extrapolation techniques. All variables were standardized using Z-score normalization.