The systems in which missing values (NULL) occur are called incomplete information systems and computations on these may lead to biased conclusions. The structured difference of the datasets and importance of attributes compels us to depend on uncertainty-based approaches for finding the null values. This paper presents a hybrid approach for solving null value problems using the concepts of rough set theory and neural network. In this, complete tuple set is used for training the NN. The incomplete tuples are then tested using the model. Level of dependency is used to judge the importance of association rules [11]. Testing the dataset after reducing unwanted attributes, yields a reduced error percentage. The system produces result with better efficiency as observed by the values of accuracy, completeness, and coverage. Thus, the proposed algorithm can be suitably modified for different scenarios using the algorithm step-by-step to solve the null value problem. © Springer Nature Singapore Pte Ltd. 2017.