Forecasting impulse buying behaviour: a comparative study of select five statistical methods

Globally, marketers and retailers believe that buying decisions are mostly made inside the store. To capture shoppers' attention and encourage such impulse buying, retailers spend heavily on in-store promotion, product display and store environment, etc. Via this study, the accuracy in predicting impulse buying behaviour using five modelling techniques and utilising R-3.2.1 as a statistical tool, has been tested with data drawn from two major Indian cities. The main objective here is to analyse and compare the capacity of various predicting methods that can be used by retailers and marketers for forecasting sales. Analysis of the data confirmed the predictive ability of the methods, albeit with a varying levels of accuracy. The study provides statistical evidence that support vector machines (SVM) significantly outperforms logistic regression, linear discriminant analysis, quadratic discriminant analysis and k-nearest neighbour methods in terms of predicting power. The findings offer a number of implications for retailers and marketers. The managerial implications of the study along with scope of further research have been addressed.