Selecting of best resume applicants from multiple sets of resumes is a tedious, labour and resource-intensive procedure. The most popular problem is that when screening resumes, there is no interpreted data that accurately labels the data they capture. This study shows the extraction of feature and then based on extraction clustering process for resume will be made. After screening the resume will be rank in order to solve such problems. To complete this work few model is developed like deep learning (DL) based Pyramid Dilated CNN with Bidirectional GRU (PDCNN-Bi-GRU) model presented to excerpt skill-based attributes from the resumes. Additionally, DL technique is hybridized which generates skill-attributes information as an output. Next the fuzzy matching module is designed to compare skill-attribute with different job categories that can enhance accuracy metrics. Candidate resumes with stronger skill sets are then categorized through an efficient hummingbird optimization method by Spectral Clustering (SCHO). The proposed method is developed and tested on both public datasets and real-time datasets using the Python platform. Performance of model is assessed using various metrics like accuracy, sensitivity, time complexity. Also, the metric specificity, kappa, and a convergence curve are evaluated with comparisons against existing techniques on both datasets. The technique achieves an accuracy of 99.3% on public datasets and 99.83% on real-time datasets, demonstrating its effectiveness.