Usefulness of Artificial Neural Networks and Experimental Measurements for the Prediction of Electrical Properties and Thermal Sensor Applications of Pr _0.8 K _{0.2– <i>x</i>} Na _<i>x</i> MnO ₃

An artificial neural network (ANN) model based on Bayesian regularization (BR) was developed to predict the electrical conductivity σ (T, x) of Pr0.8K0.2–xNaxMnO3 ceramics (with x = 0.00, 0.05, 0.10, 0.15, and 0.20) based on experimental data. The model was trained to evaluate the conductivity behavior in detail, demonstrating highly accurate and fast predictions. The ANN approach significantly accelerates material characterization by reducing the experimental workload and time. The best prediction accuracy achieved was 99.95%, with a root mean square error (RMSE) of 7.9817 × 10–7, a mean absolute error (MAE) of 0.0379 S/m, and a mean absolute percentage error (MAPE) of 1.30% at 1000 training epochs. The trained model was then used to forecast the electrical behavior at finer sodium concentration intervals (1%) and with higher temperature resolution (2 K steps up to 500 K, compared with the original 20 K steps). Finally, ANN predictions were used to identify suitable materials for thermal sensors and related applications.