The prevalence of gestating moms from various countries and ethnic groups worldwide who have gestational diabetes mellitus (GDM), a disorder characterised by abnormally high blood glucose levels, has rapidly increased. This research propose novel technique in gestational diabetes detection using machine learning technique in big data with data mining analytics. Here the input has been collected as data of pregnant women for diabetes prediction. This data has been processed for dimensionality reduction and normalization. Then it has been segmented and feature fused using attention mechanism based weighted convolutional neural networks. The experimental analysis has been carried out in terms of accuracy, precision, recall, F-1 score and AUC. Proposed technique attained accuracy of 96%, precision of 92%, recall of 85% and F_1 score of 89%, AUC of 71%.