Optimizing Consumer Insights: Predicting Customer Purchase Behavior by Using Artificial Neural Networks Approach

This study investigates the use of artificial neural networks (ANN) to predict consumer purchase behavior based on behavioral and demographic characteristics. The proposed ANN model comprises input, hidden, and output layers, optimized through hyperparameter tuning for improved prediction accuracy. The model employs non-linear activation functions, including ReLU, Sigmoid, and Tanh, alongside both Stochastic Gradient Descent (SGD) and Adam optimizers to enhance learning efficiency. Results indicate that Adam and RMSprop optimizers achieve the best balance, with Adam providing the highest accuracy (0.88) and RMSprop showing the lowest loss (0.2614) with strong accuracy (0.87), making these optimizers well-suited for capturing complex consumer behavior patterns. Further accuracy, loss, confusion matrix, ROC curve, and area under the curve (AUC)-demonstrate the model's capacity to capture intricate patterns. The confusion matrix highlights high precision in classification, with an overall accuracy of correctly predicting 262 out of 300 cases. The ROC curve further supports the model's classification strength, achieving an AUC of 0.92, which confirms its high discriminatory ability. These findings suggest the ANN model is effective for deriving consumer insights, capable of delivering accurate forecasts for data-driven marketing strategies and customer behavior optimization in practical business settings.