In this paper, we present a data-based monitoring approach designed for industrial data classification, aiming to minimize misclassifications of normal operations and to maximize the detection of anomalies and outliers. We make use of movinghorizon approaches and regression methods. Through evaluation of various algorithms on an industrial dataset, we showcase the effectiveness of the classification. As per our findings, effective detection can only be realized in conjunction of moving-horizon estimator with a regression model trained on historical measurements. The best prediction models consistently achieve accurate detection within the approved process tolerance, highlighting the efficacy of the proposed approach.