In numerous industrial sectors, data-driven machinery system condition diagnosis is imperative for maintaining machine performance and safety. To achieve this, it is crucial to analyze data patterns and accurately discern the state of the machinery. Scatter plots analysis is a valuable tool for visually understanding relationships between features and is frequently employed in data pattern analysis. However, in complex systems with a multitude of features, generating and analyzing scatter plots for all variable pairs is practically unfeasible. In this study, we propose an algorithm for automatically extracting significant Scatter plots based on Mahalanobis Distance (MD). This algorithm also facilitates the selection of pertinent features. Through this approach, we anticipate efficient identification of crucial feature factors during the data analysis process, ensuring consistency and accuracy in feature selection. Furthermore, we anticipate that this method will contribute to accurately diagnosing the state of machinery systems and optimizing their performance.