Machine learning algorithm applications still control Internet trade with their seemingly endless options for customization. Great fast data is continuously passed through socially important forecasts to improve online shopping. In the absence of analytical instruments to manage homogeneous data sets and outlines, unforeseen occurrences of data known as imbalanced data are unfortunately still overlooked. Rare cases of substance use are therefore still ignored, causing costly losses or even tragic circumstances. A number of methods have been successfully implemented to meet this challenge over the past 10 years. In many cases, however, there are significant disadvantages due to the non-uniformity of the relevant data when used for diverse application domains.