Recursive Feature Elimination based Multi-variate Naïve Bayes Classification for Product Recommendation
Bidare Divakarachari Parameshachari, S. G. Gollagi, Piyush Kumar Pareek, Mohandas Karamthoti · 2022
Today's most important E-commerce apps are those that use the latest AI and machine learning technologies. Unlike humans, computers are unable to comprehend non-digital alternatives. Therefore, it is vital that consumers learn about things like food, clothing, and medicine. Human perceptions have certain limits, but expanding them is no easy feat. Because of the rapid development has quickly become a common way for people to make purchases and get content. Performing an efficient way to increase consumer happiness. Because of shifts in sequence length, textual order, and complex logic, imagining the true polarity of user evaluations remains a formidable challenge. It is recommended that SA work be broken down into four distinct phases: I data gathering (DC), (ii) pre-processing, (iii) features extraction (FE) or term weighting-(TW), (iv) feature-selection-(FS), and (v) polarity or sentiment classifications (SC). To begin, the data collection process begins with the Web Arguing Tool (WST) being used to scrape product reviews from E-commerce sites. The data acquired from the web scrape is then preprocessed. These cleaned data are further processed using Term Frequency-based inverse class frequency by passing them via TW and FS. The purpose of this study is to use two common Naive Bayes approaches—Classification—to determine if a review of a product is positive or negative. The optimal candidate feature subset is chosen by using a recursive feature elimination strategy on the data. The study also seeks to determine which of the two methods is more effective for the provided data. A ‘2’ yardstick dataset is used for comparing the proposed classifier to the current state of the art. Consequences show that, associated to state-of-the-art procedures, the suggested model provides the highest performance in SC