How to Cite: Anima, K., & Kumar, P. S. (2022). Improvised DBSC and transductive SVM for hybrid recommendation systems. International Journal of Health Sciences, 6(S4), 4304–4317. https://doi.org/10.53730/ijhs.v6nS4.9308 International Journal of Health Sciences ISSN 2550-6978 E-ISSN 2550-696X © 2022. Manuscript submitted: 27 March 2022, Manuscript revised: 18 May 2022, Accepted for publication: 9 June 2022 4304 Improvised DBSC and transductive SVM for hybrid recommendation systems K. Anima Research Scholar, Department of Computer Science,AJK College of Arts and Science, Coimbatore, Tamilnadu - 641105, India Corresponding author email: animaabhi@gmail.com Dr P. Senthil Kumar Assistant Professor, Department of Computer Science,AJK College of Arts and Science, Coimbatore, Tamilnadu - 641105, India senthilkumar@ajkcas.com Abstract---Recommendation Systems play a crucial role in improving the business strategies and providing the customers or users with the choice to opt for a product or service. Given the growth and power of information analytics tools and inclination of data mining techniques, a wide range of recommender systems particularly for the web based users have been evolved over the decade. The conventional “One model – Fit for all” model fails in most of the cases due to the common fact that not all users have same interests and have paved way for tailor made recommender systems. It is hence highly important to provide a personalized framework which can accumulate specific interests of the users. Web data are normally highly sparse in nature and formulating a personalized user profile was a difficult task. The previous methods used user based CF method to address the former issue while Extended Jaccard Similarity (EJS) was introduced to address the latter. Although these methods found to be effective, there were scopes for improvement in the Query processing time and accuracy. To solve this problem the proposed system design an Enhanced DBSCAN and TSVM based hybrid recommendation scheme. In the proposed system, the user interests and profile are analyzed by mining the web logs. After de-noising, user profile clustering is done through improved DB scan where Anarchic Society Optimization is used for parameter fine tuning. The sequential pattern of the users is evolved based on the GSP approach. The user interests are then evaluated using the TSVM which makes use of the weighted mean for the calculation of trust value of the users. Lastly, the personalized recommendations are generated based on the clicks, selections and browsing patterns. The proposed method is applied to MOVIE LENS data and found to perform better than the previous methods and