*Corresponding Author: saurabhgyangit@gmail.com 130 DOI: https://doi.org/10.52756/ijerr.2024.v41spl.011 Int. J. Exp. Res. Rev., Vol. 41: 130-145 (2024) Optimizing User Satisfaction in Movie Recommendations Using Variable Learning Rates and Dynamic Neighborhood Functions in SOMs Saurabh Sharma 1* , Ghanshyam Prasad Dubey 1 and Harish Kumar Shakya 2 1 Department of Computer Science and Engineering, Amity University, Gwalior, Madhya Pradesh, India; 2 Department of Artificial Intelligence & Machine Learning, Manipal University, Jaipur, Rajasthan, India E-mail/Orcid Id: SS, saurabhgyangit@gmail.com, https://orcid.org/0000-0002-0090-6043; GPD, gpdubey@gwa.amity.edu, https://orcid.org/0000-0003-0868-4219; HKS, harish.shakya@jaipur.manipal.edu, https://orcid.org/0000-0002-5401-3507 Introduction Movie recommendation systems have significantly advanced through these developments. These systems can generate highly personalized recommendations by incorporating deep learning and sentiment analysis (Alatrashand Priyadarshini, 2023; Alatrash et al., 2021). Hybrid models that merge collaborative filtering and content-based strategies have effectively mitigated the cold start issue, thereby improving recommendation relevance even for new users (Li et al., 2022; Otter et al., 2022). Additionally, sentiment-enhanced collaborative filtering has been applied to enhance recommendation systems, especially in e-learning (Alatrash et al., 2021; Nain, 2023). The integration of artificial intelligence (AI) in recommender systems has broadened their scope beyond traditional applications, extending to areas like e- commerce, healthcare, and urban management. AI-driven sentiment analysis is critical in improving recommendation accuracy by comprehending user emotions and preferences (Alatrashand Priyadarshini, 2023; Mehfoozaand Basha, 2021). Research has shown the Article History: Received: 29 th May, 2024 Accepted: 21 st July, 2024 Published: 30 th July, 2024 Abstract: Customized movie recommendations are crucial in elevating user satisfaction and engagement in the era of vast online entertainment options. This study presents an innovative approach utilizing Enhanced Self-Organizing Maps (SOMs) for movie categorization. SOMs, as unsupervised neural networks, are highly effective in recommendation systems due to their ability to identify intricate data patterns accurately. The proposed method involves collecting user-movie interaction data, such as user ratings and movie attributes. Data standardization is performed to ensure consistency before training the refined SOM. By integrating variable learning rates and dynamic Neighborhood functions, the advanced SOM can uncover complex patterns within datasets, thus enhancing the accuracy of personalized movie recommendations by identifying meaningful connections between users and films. To further improve recommendation quality, hybrid filtering techniques are employed, combining content-based filtering, which considers movie characteristics like genre and description, with collaborative filtering algorithms that analyze user-item interactions to expand the range of recommended films. This integrated approach allows for the generation of user- movie matrices by employing SVD collaborative filtering to give precedence to movie recommendations. The hybrid technique demonstrates superior performance compared to earlier models, attaining an RMSE of 0.410, MAE of 0.211, precision of 92.09%, recall of 93.12%, and an F1-score of 92.15%, consequently offering very accurate movie recommendations. Subsequent studies could concentrate on improving personalised recommendations by integrating supplementary contextual data. Keywords: Film Recommendation Systems, Hybrid Recommendation System, Over the Top (OTT) Platforms, Predictive Analysis, Sentiment Analysis, User Preferences How to cite this Article: Saurabh Sharma, Ghanshyam Prasad Dubey and Harish Kumar Shakya (2024). Optimizing User Satisfaction in Movie Recommendations Using Variable Learning Rates and Dynamic Neighborhood Functions in SOMs. International Journal of Experimental Research and Review, 41(spl.), 130-145. DOI: https://doi.org/10.52756/ijerr.2024.v41spl.011