REVEALING HIDDEN TRENDS: INVESTIGATING PRODUCT SALES PATTERNS WITH CATEGORICAL AND CONTINUOUS PREDICTORS IN A DISTINCTIVE DATASET

Abstract

This comprehensive research delves deeply into the intricate web of variables that influence product sales performance within e-commerce. At the heart of our study lies a distinctive dataset meticulously curated from data.world.com, a repository that boasts a rich tapestry of 43 columns and 1,573 rows, each offering a snapshot into the diverse array of products available on the esteemed Wish.com platform. It is essential to underscore that this repository, in contrast to traditional datasets, not only comprises product listings but also intricately weaves in product ratings and sales performance metrics, thus conferring a singular perspective that ignites novel avenues for analysis. Our research journey unfurls as we deftly construct a predictive model that unveils the hidden tapestry of correlations and patterns beneath the surface of product success. With a deft interplay of categorical and continuous predictors, we undertake the task of untangling the intricate associations deeply embedded within the dataset’s fabric. Here, our ensemble of five categorical variables assumes center stage, each sentinel to fulfill specific prerequisites within a given record. This chorus of categorical variables harmonizes with six numerical features, their collective symphony orchestrated to predict, with remarkable precision, the number of units that will find eager homes. The orchestration of meaningful insights rests firmly in the capable hands of the R programming language, a formidable ally in our endeavor to analyze and assess our treasure trove of data meticulously. Our modeling odyssey reaches its zenith in forming a distilled iteration, where two categorical predictors, their symbiotic interaction, and two continuous predictors merge into a harmonious whole. With the scaffolding of linear regression, we erect a robust mathematical foundation that systematically explores the intricate dance between predictors and the response variable. A symphony of meticulous tests, encompassing individual t-tests and hypothesis evaluations, becomes the crucible in which we forge the significance of our predictors. In this crucible, we lay bare the undeniable sway of certain variables over product sales while others offer glimpses of more muted predictive power. Our discerning gaze extends to the determination of beta coefficients, confidence intervals, and the broader evaluation of model significance, each thread woven intricately into the fabric of our research narrative. In this journey, our scrutiny takes us through the labyrinthine alleys of an interaction term, and its role is dissected with utmost rigor through the prism of ANOVA and hypothesis testing. The mosaic of emerging statistical evidence compels us towards a reasonable simplification, a decision informed by the realization that its contribution to explanatory power is akin to a fleeting whisper. In summation, our study embarks on a voyage to demystify the intricate choreography that underpins e-commerce product sales. We unpick the skeins of association that weave through the constellation of predictors and sales performance, ultimately furnishing practitioners and researchers with a unique vantage point. Armed with these insights, they traverse the ever-evolving landscape of online retail with an enhanced ability to chart courses, optimize strategies, and make informed decisions that resonate with the symphony of success.