The secondary cooling water flow and the cast slab surface temperature can directly affect the quality of the slab. To address the above problems, the group has developed a wavelet weight-twin support vector machine (WTWTSVR) based model for predicting the secondary cooling water flow and the cast slab surface temperature. Compared with traditional machine learning algorithms, the model has the advantages of strong generalization ability, fast computation speed, and good fitting effect, effectively overcoming the disadvantages of multiple locally optimal solutions, equal weights, and easy overfitting that may exist in traditional neural network forecasting algorithms. To reduce the influence of noise interference on the actual data, a wavelet transform-based weighted algorithm (WTW) is introduced on the basis of a twin support vector machine algorithm, and the data is processed by noise reduction using wavelet analysis, while the reconstructed signal is given weights to avoid the overfitting phenomenon. The production practice shows that the prediction accuracy of the model is 92.6% and 93.5% for 200 furnace times (200 groups of samples); the prediction accuracy is 96.2% and 96.3% for secondary cooling water flow rate and casting slab surface temperature within the error range of ±5% and ±10%, respectively. The double hit rate within different error ranges within 10 °C reached 90.4%, which can fully meet the actual production requirements of a steel mill, and also provide guidance for the forecast of the flow rate of the secondary cooling water and the cast slab surface temperature.