QSAR modelling study of the bioconcentration factor and toxicity of organic compounds to aquatic organisms using machine learning and ensemble methods

Incremental Learning in Modelling Process Analysis Technology (PAT)—An Important Tool in the Measuring and Control Circuit on the Way to the Smart Factory

“…Current computer-aided tools have made strides towards accurate and efficient detection of chemical concentrations. However, a common drawback observed in these approaches was the use of models that suffer from unlearning [ 5 , 8 ]. In these models, the previously learned knowledge was discarded, and new models had to be trained from scratch in the learning phase as soon as new data became available.…”

“…SVM are robust classifiers, but large datasets lead to long computation times, high memory requirements, and increased complexity of the model. To solve this issue, SVM ensembles, where each SVM sees only a fraction of the data, are a viable solution [ 5 ]. Standard methods are used to achieve a high accuracy classifier by computing the best hyperparameters for the SVM model like tenfold cross-validation and grid search [ 6 , 7 ].…”

“…Generally, the studies using SVM learning from new data involve discarding the existing classifier, integrating the new data to the old set, and training a new classifier from scratch. The studies do not learn incrementally with the addition of new data, and they result in unlearning of data [ 5 , 8 ]. This means that the system cannot learn new information without forgetting previously learned classifiers.…”

Relación cuantitativa estructura actividad del factor de bioconcentración de los bifenilos policlorados en especies de peces utilizando métodos basados en aprendizaje de máquina

“…En los últimos años, los bifenilos policlorados (PCBs) han sido estudiados por los investigadores debido a sus repercusiones negativas en la salud debido a su incremento acelerado en el ambiente [1][2][3]. Los PCBs son un grupo de contaminantes orgánicos cuya estabilidad fisicoquímica les permiten resistir en el medio ambiente por un largo tiempo incluso en diferentes condiciones ambientales [3][4][5][6][7][8].…”

“…Sin embargo, la determinación del BCF a través de procedimientos experimentales presenta costos altos en cuanto a tiempo y dinero [4,8,11]. En consecuencia, algunos estudios teóricos han sido llevados a cabo para predecir características toxicológicas de las moléculas a través de modelos de relación cuantitativa estructura-actividad (RCEA) [2].…”