Tóm tắt

Background: The concept of solubility parameters was originally used for simple liquid mixtures, and later expanded to solid/liquid systems. It is useful in the cosmetics, pharmacy, coating and painting industry and also useful in determining biological material. This study presents an alternative computational method for estimating solubility parameters based on the molecular descriptor. A comparison to the traditional method is also evaluated. Materials and Methods: Neural networks (INForm) and Neuro-fuzzy logic (FormRules) were applied for predicting the value of solubility parameters based on the molecular descriptor taken from published results. Results: The artificial intelligence technique applied in this research has been introduced as a robust and highly accurate intelligent prediction technique, likely to be well suited to the solubility parameter. Neural network and neuro-fuzzy techniques achieve similar performances, but when compared to the statistical method, these techniques produced superior results for the solubility parameter. In addition, it was also illustrated that the neuro-fuzzy technique showed the relationship between molecular descriptors and solubility parameter in the form of IF-THEN rule. This becomes increasingly significant when investigating the d value on large numbers of compounds. Conclusions: The preliminary evidence presented in this paper suggests that the improved models are significant when using the approach of artificial intelligence techniques. Overall, the artificial intelligence methods employed in this research could be considered as an efficient and effective alternative approach for predicting solubility parameters.

Từ khoá

Toàn văn

PDF

Trong số này