Analysis of repurchase intentions of e-commerce consumers for sustainable business strategy build-up using a neural network-based decision support system

The rapid growth of e-commerce has transformed global commerce, providing consumers with convenience and choice while reshaping business operations. However, challenges persist, including cultivating repeat purchases and addressing concerns about local businesses. Data from 438 online consumers were collected through an online survey, with preprocessing involving outlier detection, missing value handling, and feature selection. This study classified factors affecting repurchase intentions among online shops using artificial neural networks (ANN) and compared it with random forest, XGBoost, and LightGBM. The findings revealed that the Levenberg-Marquardt Algorithm, particularly when implemented with an 85:5:10 ratio and 10 hidden layers, consistently demonstrated strong predictive performance, supported by high R 2 values. Additionally, the LSTM model exhibited a low root mean square error, indicating accurate predictions of repurchase likelihood, ranging from 4 to 4.5. The accuracy and reliability of the predictive models contributed to understanding consumer behavior in online shopping where businesses can utilize these insights to enhance customer retention and sustainable growth. Results implicated how reviews and ratings, as well as purchasing behavior and satisfaction were significant variables affecting repurchase intention. It could be deduced that the importance of perceived authenticity, credibility, and relevance of online feedback from other customers; security of transactions, assurance of high-quality products, and effective customer service also plays a pivotal role in influencing their decision-making process. The study’s findings could be applied across industries, emphasizing ongoing research and innovation to navigate the ever-changing online commerce landscape.