Nowadays, Event based Social Network has become popular flexible platform to share knowledge of various social and technical aspect among the interesting participants in the social network. Due to wide number of the advantages in event based social network , many event organizer generating multiple events across social network which increase the propagation of multiple similar events and leads to user interaction issues in predicting the successive event among competing events. Most traditional approach employed to event popularity analysis utilizes machine learning and deep learning models only on intrinsic and extrinsic properties of the user attention on various context of the event instead of focusing on evolving factors of the user. However dual attention on event and user has not be focused at same time in predicting the successive events among the competitive events to the selective group of the participants. In this paper, Online deep evolving dual attention network for successive event prediction and recommendation to target evolving participant deep is proposed to exploit the popular event to the evolving user group. Initially, dataset containing attribute with missing value has been filled using imputation method and singular value decomposition method is employed to irrelevant attribute reduction. Pre-processed dataset is employed to Latent Dirichlet Allocation towards user and event profiling to obtain the latent information of the user and event in form of feature vector. Next, Feature extraction technique considered as linear discriminant analysis is employed to extract the evolving user feature and event features containing attributes with respect to the scatter matrix. Evolving event features and user features is projected to the online deep dual attention network to compute the successive event to the user. It is carried out on processing the user attention layer and event attention layers and concatenation of the layers to yield a representation learning. It represent the mapping of the user to the event on basis of drift in both user and event feature vector on multifaceted attribute information. Finally event recommendation is provided to evolving user and participant recommendation is provided to the successive event on competition events. Evaluation of the proposed model through various case studies has been implemented and validated across various measures such as accuracy on precision, Recall and f measure along scalability and Execution time.