Predicting repurchase behavior and optimizing marketing for e-commerce users with genetic algorithms and deep...

Area of Science:

• * E-commerce Analytics
• * Machine Learning
• * Computational Intelligence

Background:

• * Increasing e-commerce competition necessitates advanced strategies for customer retention and repurchase prediction.
• * Traditional Deep Learning (DL) models often struggle to capture complex customer behaviors and optimize marketing efforts.
• * Predicting user repurchase behavior is crucial for sustainable e-commerce business success.

Purpose of the Study:

• * To develop a hybrid framework combining DL and Genetic Algorithms (GA) for accurate e-commerce repurchase prediction.
• * To optimize personalized marketing interventions for improved e-commerce strategies.
• * To enhance precision marketing and maximize customer lifetime value.

Main Methods:

• * Data preprocessing involved handling missing values, removing duplicates, and min-max normalization.
• * A Dynamic Genetic Algorithm-mutated Enhanced Vanilla Memory Network (DGA-EVMN) was utilized for predicting repurchase possibilities.
• * A Dynamic Genetic Algorithm (DGA) was applied to optimize personalized marketing campaigns.

Main Results:

• * The DGA-EVMN framework demonstrated superior performance over baseline models.
• * Achieved high prediction metrics: F1-score (0.960), AUC-ROC (0.971), recall (0.962), precision (0.958), and accuracy (0.984).
• * DGA-driven optimization led to efficient resource allocation, reduced marketing waste, and increased conversion rates.

Conclusions:

• * Combining predictive analytics with evolutionary optimization offers a powerful approach for intelligent marketing.
• * The proposed DGA-EVMN framework significantly improves e-commerce repurchase prediction and marketing effectiveness.
• * This data-driven strategy enables adaptive and cost-effective practices to maximize customer lifetime value.