| dc.description.abstract | The COVID-19 pandemic has significantly accelerated the growth of e-commerce
globally, necessitating businesses to improve customer satisfaction to remain
competitive. Customer satisfaction is a significant factor deciding long-term success of
a business, predicting customer satisfaction can help businesses plan their strategies and
tackle the problems. This thesis focuses on predicting customer satisfaction in e commerce, using a dataset provided by Olist e-commerce company in Brazil which
dated between 2016 and 2018, containing information about orders, products, sellers,
and customers. The study aims to build a predictive model employing machine learning
techniques, customer satisfaction levels are classified into 2 classes: satisfied and not
satisfied. The methodology involves data preprocessing, feature engineering to modify
the dataset to make it suitable and reliable for running models. The used predictive
models were logistic regression, decision tree and random forest. Logistics regression
was found to be best model with value of F1-score of 92.08%. The results suggest that
machine learning can effectively predict customer satisfaction which can be used for
future applications. | en_US |
