Today, fraudulent credit card transaction are committed by unauthorized individuals and organization exploring methods such as phishing and social engineering fraud. Several researchers proposed Machine Learning (ML) techniques to serve as solutions to credit card fraud detection. However, ML suffers imbalance class distribution, high dimensionality and sparsity problem. These concerns make it extremely difficult in credit card fraud detection. Therefore, this study proposes Deep Convolutional Neural Network (DCNN) with Synthetic Minority Oversampling Techniques (SMOTE) to serve as an ideal solution. Data augmentation method of SMOTE oversampling techniques is incorporated into the DCNN that solve the class imbalance trial and in improving the performance. Datasets with 284807 records with 31 features gotten via Kaggle repository is used. Implementation is carried out on Google Colab containing cloud-based platform embedding Jupyter notebook environment with Graphical processing unit (GPUs). Two experiment were carried out where the first experiment is used to determine suitable models among baseline techniques: Logistics Regression (LR), Random Forest (RF), Isolation forest, and single DL model of Multiple layer perceptron (MLP). The baseline models yielded an overfitting accuracy score, with recall, specificity, precision, and F1-score all presenting 1.00% respectively. This outcome is not sufficient in establishing findings on imbalance data distribution as its biased. This led to the construction of a new ML model absorbing Light Gradient Boosting Machine (LGBM), with Artificial Neural Network (ANN) and proposed DCNN+SMOTE for the second experiment phase alongside baseline models. Experimental results via simulation show the proposed DCNN+SMOTE yielded overwhelming superclass performance across board displaying 1.00% results respectively better than baseline models. It Error Rate (ER) and Null Error Rate (NER) is 0.00% distinctly. While, the False Positive Rate (FPR) yields 0.001% result lesser than the baseline models.