Paramedical science is the developing strategies in the medical field for growing drug patterns to create new com-pound molecules. The compound molecules are non-relation to suggest the drug to the patients to cause side effects. This accuses of compound molecule features being non-depended with each other doesn't create patterns to clas-sification inaccuracy. To resolve this problem, we propose a relative drug compound analysis based on Radial Sub-set Clustering Feature-Based Deep Spectral Neural Classification (RSCF-DSNC) for predicting relational drug rec-ommendation. Initially, the dataset is collected from paramedical compound dataset related to paracetamol, and diphenhydramine hydrochloride bioinformatics data are pre-processed. Then the featured labels are marginalized to estimate the Relative Drug Intensive Weight (RDIW) for finding the mutual correlation. Based on the correlation weight, the Radial Subset Clustering Feature Selection (RSCFS) is applied to predict the relative closeness of drug molecules. Infra Segment Subset Feature Pattern Theory (ISSFPT) technique is used to analyze the successive pat-tern weight from the RSCFS dataset. Then the selective feature patterns are then trained into Deep Spectral Neural Classification (DSNC) adapted with Convolution Neural Network (CNN) to identify the relational class of drug mol-ecules category. The proposed drug success rate prediction performance result is 93% low time complexity 31sec. This produces high prediction accuracy compared to the other system than other methods for recommending fea-tures for relation to drug compound recommendation.

Drug compound analysis, pattern prediction, subset clustering, feature selection