Study of Nematic Transition Temperatures in Themotropic Liquid Crystal Using Heuristic Method and Radial Basis Function Neural Networks and Support Vector Machine

Abstract

Quantitative Structure -Property Relationships (QSPRs) models have been successfully developed for the prediction of the nematic transition temperatures (T N ) of 42 thermotropic liquid crystals. Heuristic Method (HM) and Radial Basis Function Neural Networks (RBFNNs) and Support Vector Machine (SVM) were utilized to construct the linear and non-linear QSPRs models, respectively. Comparing the whole results obtained from the three models, the RBFNNs model was much better. The optimal QSPRs model which was established based on RBFNNs gave a square correlation coefficient (R 2 ) of 0.984, 0.953, 0.973, and Root-Mean Square (RMS) error of 2.19, 4.13, and 2.99 for the training set, the test set, and the whole set, respectively. Some analysis to the dataset and evaluation were done in the paper. All the results indicated that the proposed QSPRs model was robust and satisfactory.

The Applications of Machine Learning in the Study of Liquid Crystals: A Review

“…A Multi-linear regression model was able to calculate the nematic transition temperatures of LCs with two aromatic rings linked to an ester group with varying terminal chains using six descriptors calculated using DFT, which proved to be comparable to other MLR models using descriptors calculated by different methods (Al-Fahmi, 2014). Based on eight different descriptors of molecules that all shared a COO-chain linking two terminal hydrocarbon chain, a Radialbased function neural network, a feedforward neural network with only one hidden layer and a radial-based function (whose value at any point is solely dependent from the distance from the input to some other point) for an activation function in the hidden layer (Karayiannis and Mi, 1997), was able to predict the phase transition temperature for these types of molecules with an R-squared value of 0.953, performing better on validation data than the Support Vector Machine model trained on the exact same data (Gong et al, 2008). Using six descriptors of LCs with the exact same specifications as Al-Fahmi's study, both MLR and feedforward neural networks were tested to predict the nematicisotropic transition temperature, with the neural network performing better on both training and validation sets than the MLR model (Fatemi and Ghorbanzand'e, 2009).…”

A GMDH-type neural network with multi-filter feature selection for the prediction of transition temperatures of bent-core liquid crystals

“…5 Quantitative structure-property relationship (QSPR) methodology has been oen used to predict various physical and chemical properties of LCs. [4][5][6][7][8][9][10] Articial neural networks (ANNs), as a nonlinear modelling approach, are mostly used for this purpose, due to complex relationships exist between a property of molecule and its structure. 6 Among rst, Johnson and Jurs 4 have shown that the clearing temperatures of a series of structurally similar rod-like LCs can be successfully predicted using ANNs.…”

“…[4][5][6][7][8][9][10] Articial neural networks (ANNs), as a nonlinear modelling approach, are mostly used for this purpose, due to complex relationships exist between a property of molecule and its structure. 6 Among rst, Johnson and Jurs 4 have shown that the clearing temperatures of a series of structurally similar rod-like LCs can be successfully predicted using ANNs. In a recent study, Antanasijević et al have used QSPR method in combination with ANNs, decision trees (DTs) and MARS (multivariate adaptive regression splines) technique for the prediction of liquid crystallinity, 10 and with DT and MARS for the estimation of the clearing temperatures 7 of vering bent-core molecules.…”

Prediction of liquid‐crystalline property using support vector machine classification

“…Gong et al . applied heuristic method, RBFNN, and SVM for the prediction of nematic transition temperature.…”

“…The results provided by the neural model were far better than the previous ones obtained with the MLR model used by Villanueva-Garcia et al [29] for the same experimental database. Gong et al [30] applied heuristic method, RBFNN, and SVM for the prediction of nematic transition temperature.…”