The primary objective of this work is to discover a meaningful information in heart disease dataset for better diagnosis. This work is done using the data set available in UCI Machine learning repository. The work focuses on selecting the important features in the dataset using Principal Component Analysis and regression techniques. Using regression, the exponentiated estimate of the coefficient exp(B) of the feature is considered for feature selection. The exp(B) is the odds ratio of the independent variables. The work is done taking into consideration the components extracted using Principal Components Analysis technique and applying various operations on these components to produce methods like PCA1, PCA2, PCA3 and PCA4. It is observed that for one of the proposed methods PCA1, the prediction accuracy is 92.0{\%} using regression and 95.2{\%} using feed forward neural network classifier which is better than other methods. It is also observed that the accuracy of exp(B) is closer to PCA1 method, hence concluding that the exp(B) can also be considered for feature selection. {\textcopyright} 2013 Springer International Publishing.