Measuring E Customer Satisfaction in Online Retail A Comparative Analysis Using Various ML Algorithms

Maintaining competitive advantage in the often-shifting arena of online retail increasingly rests on knowing and increasing consumer happiness. As online buying becoming increasingly frequent, retail companies are drowning in massive amounts of consumer feedback data. Maximizing service offerings and enhancing user experience depend on knowledge of consumer satisfaction derived from this data. This paper looks at how different ML methods could fairly assess consumer feedback and project degrees of satisfaction. The primary challenge is exactly gauging consumer happiness from unstructured input data. Two areas where conventional methods frequently fall short are scalability and accuracy. This work underlines consumer input into satisfaction categories by means of RF, NB, and SVM algorithms under evaluation and comparison, thereby addressing this challenge. The study makes use of a dataset comprising consumer reviews and satisfaction evaluations from numerous online retail stores. First treated for missing values, the data is normalized for uniformity. For the RF, NB, and SVM algorithms, this dataset drives testing and training. Accuracy, precision, recall, F1-score, and other criteria evaluate every approach. RF achieves an accuracy of 85%, NB 78%, and SVM 82% according to preliminary numerical results. RF performs superior in terms of precision and recall especially in handling biassed datasets. Although SVM also performs pretty well, it is rather less good than RF in differentiating between closely related levels of pleasure.