Mindset reading of a student towards technology is a challenging task. The student's demographic features prediction has a significant aspect for the learning activities in educational institutions. The current studies predicted the student's native place based on technological awareness having various features such as development, availability, usability, educational benefits, etc. However,these studies have not explored the identification of sentiment identification about the technology through ML,optimization,etc.Motivated from these facts,in this paper, we propose a machine learning (ML) model with optimizing techniques to tune the hyper-parameters. In the proposed model, a primary dataset gathered from Indian and Hungarian universities, which analyzed with a Multi-Layer-Perceptron (MLP) with three popular optimization algorithms, such as Adaptive Moment Estimation (Adam), Stochastic Gradient Descent (SGD), Limited-memory Broyden-Fletcher-Goldfarb-Shanno (LBFGS). The optimized MLP has compared with the Support Vector Machine (SVM). Besides, numerous testing methods and to select the most prominent features, Principal Component Analysis (PCA) trained both models. Association of the Adam optimizer with the ReLu activation function in the MLP proved significant play in prediction with regularization. The PCA components covering most of the variance improved the optimized MLP accuracy with 2.3% and boosted the accuracy of the SVM with 2.9%. The Gain-ratio and the Info-gain suggested 11 features with significant weights. Both predictive models are found not only competitive but also outperformed with an identical prediction accuracy of 94% to identify the native place of the student. The Statistical t-test supported the equal predictive strength of both models and proved the significant enhancement in the SVM performance using the PCA components. Further, a considerable reduction is also achieved in the prediction error and prediction time to support the institute's web-based real-time system. Based on deep experiments, we recommend the optimistic native identification models for the higher educational institutions to analyze the attitude and technical awareness among students based on their native place.

No abstract

The effect of some nonylphenyl-ethylene oxide polymers on the growth of Bacillus megaterium, B. cereus vat. mycoides, B. polymyxa, B. subtilis, Pseudo-monas fluorescence and Azotobacter chroococcum was investigated in the concentration range 20-800 ppm with the agar diffusion method. The zones of inhibition, restricted growth and eventual stimulation were determined with a Shimadzu C-930 dual wavelength TLC scanner. The data matrix was evaluated by principal component analysis. A. chroococcum was insensitive to each tenside at each concentration. The growth of the other microorganisms was inhibited by the tensides. With B. megateriurn and B. cereus oar. mycoides stimulation was also observed. The effect of the non-ionic tensides decreased with increasing length of the hydrophilic ethylene oxide chain. This phenomenon can be explained by the assumption that the activity of tensides depends on their membrane-damaging effect. The bulky nonylphenyl group inserts between the apolar fatty acid chains disorganizing the membrane structure. The longer hydrophilic ethylene oxide chain modifies the distribution of tenside between the apolar and polar regions of the membrane, preferring the aqueous phase. This results in the decrease or loss of biological activity.

No abstract

An experimental study was conducted to predict the student's awareness of Information and Communication Technology (ICT) and Mobile Technology (MT) in Indian and Hungarian university's students. A primary dataset was gathered from two popular universities located in India and Hungary in the academic year 2017–2018. This paper focuses on the prediction of two major parameters from dataset such as usability and educational benefits using four machine learning classifiers multilayer perceptron (ANN), Support vector machine (SVM), K-nearest neighbor (KNN) and Discriminant (DISC). The multi-classification problem was solved with test, train and validated datasets using machine learning classifiers. One hand, feature aggregation with the train-test-validation technique improved the ANN's prediction accuracy of educational benefits for both countries. Another hand, ANN's accuracy decreases significantly in the prediction of usability. Further, SVM and ANN outperformed the KNN and the DISC in the prediction of awareness level towards ICT and MT in India and Hungary. Also, this paper reveals that the future awareness level for the educational benefits will be Very High or Moderate in both countries. Also, the awareness level is predicted as High and Moderate for usability parameter in both countries. Further, ANN and SVM accuracy and prediction time is compared with T-test at 0.05 significance level which distinguished CPU training time is taken by ANN and SVM using K-fold and Hold out method. Also, K-fold enhanced the significant prediction accuracy of SVM and ANN. the authors also used a STAC web platform to compare the accuracy datasets using T-test and ANOVA test at 0.05 significant level and we found ANN and SVM classifier has no significant difference in prediction accuracy in each dataset. Also, the authors recommend presented predictive models to be deployed as a real-time module of the institute's website for the real-time prediction of ICT & MT awareness level.