Several data mining methods have been proposed in the literature to obtain good results. However, how to effectively use and reconstruct acquired patterns is still an open research question, especially in text mining. Most modern text mining techniques primarily adopt a termbased methodology, and therefore all suffer from the problems of multiple words and synonyms. Semantic word representation is a key task in natural language processing and text mining. The knowledge of features relation has shown their influence on classification in a variety of tasks. The features relation patterns are powerful in associating features associations and effectively bridge lexical gaps, facilitating numerous applications. Most research focuses on term processing by encoding contextual information. However, many potential relationships, such as relation association patterns and semantic-conceptual relationships are not accounted for well. In this paper, we propose an integrated method known as FRP-LSRA for enhancement in information classification using Feature Relation Patterns (FRP) and Latent Semantic Relation Analysis (LSRA). The FRP is constructed using a modified Bayesian mechanism that allows discovering the relation of features. It provides the primary knowledge of the data association with their related class. The constructed patterns by FRP are being utilized for the analysis of their semantic relation to classify the data accurately. The experiment analysis is performed over an SFPD dataset to evaluate the classification enhancement, the proposed FRP-LSRA achieve 3% improved accuracy with comparing the state-of-art classification methods.