Banking fraud prevention as well as risk management are dominant in an advanced financial landscape and the integration of Artificial Intelligence (AI) provides the promising path for advancements in these areas. However, because of the nature of banking sectors inherent, data security cannot be threatened. Hence, this research proposes the Weighted Extreme Gradient Descent Boosting (WXGBoost) approach for the prediction of the attacks in banking and financial investments in cybersecurity. Initially, the data is collected from the various banks of transaction details. Then, the min-max normalization technique is utilized to normalize the collected of the transaction details from the banks. To assure the security of the financial sector management, the Advanced Encryption Standard (AES) algorithm is used for both encryption and decryption of the data. The proposed WXGBoost approach enhances the effectiveness of the cyber security through enhancing their protection over cyberattacks. The experimental results shows that the proposed WXGBoost approach attains the accuracy of 0.894, precision of 0.885, recall of 0.893 and F1-score of 0.901 respectively. These obtained values from the proposed method attains better results when compared to the existing methods K-Nearest Neighbor (KNN) and Reinforcement Learning (RL).