Appendix Data (Excel): Per-Sensor Forecasting Metrics, Signal Characterisation, and Model Hyperparameters

This dataset provides supplementary Excel tables supporting a signal-aware evaluation of seven recurrent neural network architectures for multi-zone temperature forecasting in an industrial injection moulding system. The dataset includes (i) per-sensor detailed forecasting performance metrics (MSE, MAE, MAPE, R², relative deltas, and tier classification) for 11 temperature sensors, (ii) signal characterisation metrics capturing temporal complexity, persistence, and distributional properties (permutation entropy, sample entropy, lag-1 autocorrelation, skewness, kurtosis, mean, and standard deviation), and (iii) a summary of model architectures and training hyperparameters used across all evaluated models. These supplementary data are provided to enhance transparency, reproducibility, and detailed cross-sensor comparison beyond the space limitations of the main manuscript.