Sarcomas are rare malignancies with diverse types and high prevalence in adolescents and young adults. Despite advances in treatments, sarcomas have a high mortality rate with lack of effective biomarkers. The use of biomarkers, such as microRNAs (miRs), can improve prognosis of sarcomas. Machine learning and bioinformatics tools are essential in analyzing large datasets to identify potential therapeutic targets and prognostic indicators, ultimately improving outcomes for sarcoma patients. Methods: The TCGA dataset provided miR-Seq data of sarcoma patients along with clinical parameters. Data preprocessing involved filtering and normalization steps using R programming, followed by the identification of miRs using a Deep Neural Network (DNN). Machine learning algorithms were applied for classification and feature selection in identifying predictive miR markers in sarcoma. The implementation and evaluation of machine learning models were done using Python, with performance metrics such as AUC, accuracy, F۱ Score, R۲ score, and confusion matrix used to measure the significance of miRs. The study included Kaplan-Meier survival analysis to identify prognostic markers. Results: In this study, there were ۱۱۹ females and ۱۴۲ males in the population. Out of them, ۲۵۹ individuals (۹۹.۲%) had sarcoma, ۲ (۰.۸%) were healthy, ۱۶۲ (۶۲.۱%) were still alive, and ۹۹ (۳۷.۹%) had passed away. Among the ۲۶۱ samples, a total of ۱۰۴۶ miRs were identified. The average age at diagnosis was ۶۰.۸۷ with an average follow-up period of ۸۶۳.۴۴ days. The miRs data was processed for machine learning and DNN by normalization, filtering, and feature extraction based on a correlation coefficient threshold of > ۰.۵. Subsequently, ۷۰ significant miRs related to tumor vs. normal conditions were identified. A DNN analysis showed an accuracy of ۷۹.۱۵%, MSE of ۲.۲۱, RMSE of <۰.۰۱, R۲ of ۰.۹۹, and AUC of ۱.۰ for the essential miRs. A subset of ۲۰ important miRs was identified, with ۱۰ having coefficients greater than ۰.۹۰ for further DNN investigation. Kaplan-Meier survival analysis revealed that downregulating miR.۳۱۵۰, miR.۳۶۸۰, miR.۱۲۲۹, and miR.۱۲۶۵ significantly improved overall survival in sarcoma patients, while downregulating miR.۳۱۴۴ and miR.۳۹۴۱ decreased overall survival. Statistical analysis was carried out using SPSS version ۲۰, with significance set at p< ۰.۰۵. Conclusion: Our findings highlight the need for additional research and confirmation to determine the practicality of using these miRs as potential prognostic biomarkers for sarcoma. This could potentially lead to improved management of the disease.