Due to the implicit characteristics of learning disabilities (ID), the identification or diagnosis of students with learning disabilities has long been a difficult issue. In fact, there is little consensus about what is the best procedure to identify a person with LD. Instead, the procedures are based on empirical findings from scholarly research. In this paper, we try to adopt Artificial Neural Network technique, which has been applied successfully to solve problems in numerous fields, to the ID identification and diagnosis problem. The preliminary results are satisfactory and can be provided as second opinion to the ID evaluation personnel. But it still requires many efforts to make the model more accurate and to make the idea practically feasible, which is what we will be working on.