Click traffic plays a major role for advertisers and platforms hosting content. Higher click traffic indicates greater interest in advertisements and promoted material. Compensation is usually associated with click traffic volume, with higher click rates leading to better user-to-customer conversion rates and improved sales and promotion outcomes. However, this link between click rates and compensation has led to an interesting phenomenon known as click fraud, where clicks are generated without genuine interest to exploit advertisers. This can damage the reputation of platforms, deplete advertising budgets, and impact revenue streams for content creators. Nevertheless, distinguishing genuine clicks from fraudulent activity can be challenging, as attackers constantly adapt their strategies to avoid detection. This work explores the use of click-stream analysis by applying gated recurrent unit (GRU)-based networks to identify fraudulent click traffic. Since the performance of classifiers remains tightly coupled with parameter selection, a modified metaheuristic algorithm is introduced to improve outcomes through hyperparameter tuning. Simulations on real-world data have shown promising results, with the best-performing models achieving an accuracy as high as 0.790146.