Modelling of halomethanes using neural networks

Boiling Points of Halogenated Aliphatic Compounds: A Quantitative Structure−Property Relationship for Prediction and Validation

“…Experimentally determined boiling points have been published for many of them, but there are also gaps in the available data. It is therefore of great interest that several reports have appeared regarding the estimation of the boiling point for short-chained halogenated aliphatic compounds and derivatives, either specifically or as part of a larger set of compounds. ,− The development of estimation methods is, however, always limited by the availability of reliable calibration data.…”

“…Outlier detection capabilities and adherence to the parsimony principle are both critical to the future reliability of any multivariate calibration model . Linear regression has traditionally been the method of choice for QSPR modeling, but many of the recent studies on boiling point estimation rely on neural networks. ,,,,,− Lack of interpretability is a major drawback with the more complex neural network modeling procedure, and performance has also been questioned. , Furthermore, neither traditional linear regression nor the neural networks incorporate outlier detection as a standard feature. Projection-based methods such as principal component regression (PCR) and partial-least-squares regression (PLSR) may therefore offer a promising alternate approach. , …”

“…28 Linear regression has traditionally been the method of choice for QSPR modeling, but many of the recent studies on boiling point estimation rely on neural networks. 2,8,12,18,20,[29][30][31][32] Lack of interpretability is a major drawback with the more complex neural network modeling procedure, and performance has also been questioned. 33,34 Furthermore, neither traditional linear regression nor the neural networks incorporate outlier detection as a standard feature.…”