The use of credit cards for online purchases has significantly increased in recent years, but it has also led to an increase in fraudulent activities that cost businesses and consumers billions of dollars annually. Detecting fraudulent transactions is crucial for protecting customers and maintaining the financial system's integrity. However, the number of fraudulent transactions is less than legitimate transactions, which can result in a data imbalance that affects classification performance and bias in the model evaluation results. This paper focuses on processing imbalanced data by proposing a new weighted oversampling method, wADASMO, to generate minor-class data (i.e., fraudulent transactions). The proposed method is based on the Synthetic Minority Over-sampling Technique (SMOTE), Adaptive Synthetic Sampling (ADASYN), and weight adjustment to identify specific minority areas while retaining data generalization and accurately identifying patterns associated with fraudulent transactions. Experimental results obtained from two datasets with Autoencoder (AE), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) learning models show that wADASMO surpasses other oversampling methods in three evaluation metrics: accuracy at 95.6%, 98.8%, and 99.2%; detection rate at 90.4%, 93.38%, and 93.38%; and area under the curve (AUC) at 93%, 96%, and 96.3% for AE, CNN, and LSTM models, respectively.