Predicting Purchase Intentions in Bangladesh’s Youth Market: A Competent Analysis of Machine Learning and Deep Learning Models

The rapid evolution of consumer behavior in the growing market of Bangladesh requires robust models to predict purchase intention, especially for the youth market. In our work, we assess the effectiveness of various machine learning and deep learning techniques, including linear regression, decision tree, gradient boosting, random forest, artificial neural networks (ANN), feedforward neural networks (FFNN), and recurrent neural networks (RNN). We have verified the models using important Key performance indicators such as accuracy, precision, recall, F1 score, mean squared error (MSE), mean absolute error (MAE), training time, and prediction time. Our experimental results show that RNN and FFNN provides superior predictive power for our dataset achieving the highest accuracy (99% and 98%) and better performance in other metrics. Random forest follows closely with 96% accuracy, while gradient boosting and ANN achieve 88% and 90%, respectively. Among all the models discussed, traditional ones like linear regression models scored lowest accuracy of 86% which emphasizes their shortcomings in capturing the complex relationships in the dataset. These findings highlight the superiority of ensemble and deep learning methods in handling nonlinear and high-dimensional data for accurate and robust predictions. The study ends with strategic insights for marketers and researchers in Bangladesh, emphasizing the importance of model selection for predicting purchase intention of youth demographic. Future research directions include optimizing deep learning architectures, exploring hybrid models, and enhancing real-time scalability for improved prediction accuracy.