Consumer purchase behaviour has become a potential research area in business analytics, as exploring micro-level details would increase the business's profitability. In this prospect, many MNCs and other enterprises harness contemporary computing technologies like Big Data Analytics, Deep Learning, and Predictive Analytics to explore the latent knowledge in purchase patterns and customer behaviour. This work deploys a novel Multi-class Ada Boost (MAB) supported Convolutional Neural Network (CNN) to learn customer purchase behaviour by analysing the buying patterns and trends to predict the repurchases. The proposed model learns the trends sequentially as the CNN models are cascaded one after the other, thus preserving the contextual knowledge between the models. The proposed model is tested for its efficacy on Instacart Market Basket Analysis to predict whether the customer is repurchasing the same product. The performance of the proposed model is compared with other states of art Machine Learning algorithms like Logistic Regression (LR), Support Vector Machines (SVM), Random Forest (RF), and XGBoost in terms of prediction accuracy, precision, and F1-score. In addition, synthetic noise is induced into the dataset at various levels to analyse the model's efficacy in handling noisy data. These results indicate that the model shows better results than its peers, thus making it more suitable to predict customer repurchase behaviour and pattern.