<p>The exponential growth of online shopping and other digital payment methods has led to a meteoric rise in the use of credit cards. Analyzing consumer data for the purpose of detecting and preventing fraud has been greatly facilitated by machine learning (ML). But most real-world credit card data have irrelevant and redundant features, which makes ML classifiers not work as well. In this research paper, a new approach for detecting and managing risk and segmenting customers in financial systems is introduced, called RiskNet. The proposed approach consists of three main modules, which are (i) data preprocessing module (DPM), (ii) feature selection module (FSM), and (iii) risk predictive module. For the first module (e.g., DPM), the input data are cleaned and preprocessed to be suitable for the next module. Next, during the second module (e.g., FSM) the most important and effective features are selected using the proposed feature selection method called improved whale optimization algorithm. Then, these features are fed to the used classifiers to detect credit card fraud. Finally, to assess the efficacy of RiskNet, experiments were conducted on a real-world financial dataset, and the results demonstrated that RiskNet outperforms existing state of the art. RiskNet delivers precise risk predictions and personalized risk management strategies for each customer, enhancing the overall stability of the financial system.</p>