A Computational Model to Predict Consumer Behaviour During COVID-19 Pandemic

The knowledge-based economy has drawn increasing attention recently, particularly in online shopping applications where all the transactions and consumer opinions are logged. Machine learning methods could be used to extract implicit knowledge from the logs. Industries and businesses use the knowledge to better understand the consumer behavior, and opportunities and threats correspondingly. The outbreak of coronavirus (COVID-19) pandemic has a great impact on the different aspects of our daily life, in particular, on our shopping behaviour. To predict electronic consumer behaviour could be of valuable help for managers in government, supply chain and retail industry. Although, before coronavirus pandemic we have experienced online shopping, during the disease the number of online shopping increased dramatically. Due to high speed transmission of COVID-19, we have to observe personal and social health issues such as social distancing and staying at home. These issues have direct effect on consumer behaviour in online shopping. In this paper, a prediction model is proposed to anticipate the consumers behaviour using machine learning methods. Five individual classifiers, and their ensembles with Bagging and Boosting are examined on the dataset collected from an online shopping site. The results indicate the model constructed using decision tree ensembles with Bagging achieved the best prediction of consumer behavior with the accuracy of 95.3%. In addition, correlation analysis is performed to determine the most important features influencing the volume of online purchase during coronavirus pandemic.

Cumulant-based trapezoidal basis selection for heart sound classification

Past decades witnessed the expansion of linear signal processing methods in numerous biomedical applications. However, the nonlinear behavior of biomedical signals revived the interest in nonlinear signal processing methods such as higher-order statistics, in particular higher-order cumulants (HOC). In this paper, HOC are utilized toward heart sound classification. Heart sounds are presented by wavelet packet decomposition trees. Information measures are then defined based on HOC of wavelet packet coefficients, and three basis selection methods are proposed to prune the trees and preserve the most informative nodes for feature extraction. In addition, an approach is introduced to reduce the dimensionality of the search space from the whole wavelet packet tree to a trapezoidal sub-tree of it. This approach can be recommended for signals with a short frequency range. HOC features are extracted from the coefficients of selected nodes and fed into support vector machine classifier. Experimental data is a set of 59 heart sounds from different categories: normal heart sounds, mitral regurgitation, aortic stenosis, and aortic regurgitation. The promising results achieved indicate the capabilities of HOC of wavelet packet coefficients to capture nonlinear characteristics of the heart sounds to be used for basis selection.