Now in these days the information need is increasing rapidly in our day to day life therefore a large number of users are accessing data from search engine. The search engines are composed with three major components user query interface, search algorithm and the ranking process. During search process the system evaluate the user input query and the database documents according to best fit documents are retrieved. The retrieved document is then ranked according to the user query relevance thus most near document of the user query is listed first. The available technique are provides the ranked listing of documents. In this presented work first the recently developed text document retrieval models are evaluated and then after a traditional model of document retrieval is enhanced with help of supervised classification technique. The proposed data model of the document search first finds the document’s word probability using the Bayesian classification approach then after the data is normalized to find the similar length of text document features. These document features are used to make training of neural network .The neural network processes the input training features and makes training for the documents pattern. This data model is used to predict the user input data patterns from the existing set of data. The implementation of the proposed technique is performed using the JAVA development technology after implementation of the desired document retrieval technique the performance of the system is estimated in terms of accuracy, error rate, memory consumption and the time consumption. According to the evaluated results the performance of the algorithm is found more optimum. Thus the given model is more adoptive as compared to the traditional approaches available.