In data mining approach it simply assumes a random sample of different items of a single relational database. Many data mining techniques can be used to extract information from the data. Here we have proposed the work which introduces link analysis procedure discovers relationships between relational databases or graph. This work can be useful on single relational databases as well as multiple relational databases. This approach interested to analyze two or more relational databases. By taking a random walk on the database, different states can be defined. In first part relational database is displayed in terms of graph and in second part markov chain, which contains only the interested elements and preserves the characteristics of original chain and all those elements are analyzed by projecting diffusion map. Several datasets are analyzed by using the proposed methodology showing the benefits of this technique for finding relationships of relational databases or graphs.