Issuu

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 11 Issue: 10 | Oct 2024

p-ISSN: 2395-0072

www.irjet.net

AN ENHANCING MOVIE RECOMMENDATION SYSTEM USING HYBRID MODELS R.Kirubahari1, P.Kiruthika2, S.Lakshita3, J.M.Namritha Shree4 1Associate Professor, Dept. of Computer Science and Engineering, KLN College of Engineering, Tamil Nadu, India

234UG student, Dept. of Computer Science and Engineering, KLN College of Engineering, Tamil Nadu, India ---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - The cold start problem in recommendation

To further enhance personalization, proposed model incorporates geographical location data, allowing the system to recognize region-specific movie trends, preferences, and cultural nuances that influence viewing habits. This geographical data enables the system to align recommendations with local tastes and regional popularity, improving the relevance of suggestions for users across different locations.

systems, particularly in the movie recommendation domain, arises when the system has insufficient user data, making it challenging to deliver personalized and accurate suggestions. This issue is especially significant when dealing with new users or items, which lack historical interaction data. To mitigate this problem, by integrating text analysis techniques, specifically utilizing text blob and sentiment analysis algorithms, to extract meaningful insights from the minimal user input. These algorithms help analyze initial interactions, such as user reviews or feedback, to generate relevant movie recommendations early in the user's journey, even with sparse data. Additionally, recognizing that user preferences can fluctuate depending on their emotional states, the system integrates emotion recognition techniques to refine its recommendations.

Recognizing the dynamic nature of user preferences, which may fluctuate based on emotional states, the system also integrates emotion recognition techniques. By analyzing the user’s current emotional state from textual inputs or interactions, the system can recommend movies that either align with or counterbalance their emotions. For instance, users exhibiting signs of stress or sadness may be presented with uplifting or calming movie options, creating a more personalized and emotionally attuned experience.

Key Words: Cold start problem, Recommendation systems, Sentiment analysis, User preferences, Recommendation accuracy.

1.1 PROBLEM STATEMENT The cold Start Problem in Recommendation Systems Many existing recommendation systems struggle with the cold start problem, where they are unable to generate accurate suggestions for new users or items with limited interaction data. Recommender Systems face challenges like scalability, diversity, accuracy, and data sparsity, impacting prediction quality. Tuning the hyperparameters of Restricted Boltzmann Machines (RBM) is difficult and affects performance. The Improved RBM with Bayesian Optimization (IRBM-BO) is proposed to optimize hyperparameters like learning rate, momentum, and light-cost. This method enhances prediction accuracy, verified on datasets like Movie lens and Netflix using MAE and RMSE metrics [5]. This system addresses the limitations of traditional movie recommendation systems, such as cold start, data sparsity, and lack of diversity. It proposes a hybrid method combining social similarity and item attributes to improve recommendation accuracy and relevance. This approach aims to enhance user satisfaction by providing diverse and personalized movie suggestions[1]. The proposed model aims to address this issue by utilizing hybrid models that incorporate text sentiment analysis and geospatial data, enabling the system to provide relevant recommendations even in scenarios where little user history is available.

1. INTRODUCTION The cold start problem is a prevalent challenge in recommendation systems, particularly in the movie recommendation domain, where systems struggle to provide personalized and accurate suggestions for users who lack historical interaction data. This problem is especially significant when dealing with new users or items, as the absence of sufficient data hampers the system's ability to deliver relevant recommendations. This system addresses the challenge of noise and missing data in recommendation systems, which negatively impacts recommendation accuracy. It proposes collaborative denoising auto-encoders to improve Top-N recommendations by effectively filtering out noisy information and reconstructing user-item interactions[4]. To address this issue, by integrating the text analysis techniques, specifically leveraging Text Blob and sentiment analysis algorithms, to extract valuable insights from minimal user input. By analyzing initial interactions such as user reviews or feedback, these algorithms help generate relevant movie recommendations early in the user journey, even when data is sparse.

Impact Factor value: 8.315

ISO 9001:2008 Certified Journal

Page 832