Algorithm a culturally sensitive recommender system to solve cold start problems

A fundamental problem faced by modern recommendation systems is the cold-start phenomenon, which is the inability to generate personalized recommendations when historical data on user preferences is scarce. Traditional methods of solving this problem involve collecting information through questionnaires or involving data from third-party sources, which may lead to compromising user privacy. In this paper, we propose an algorithm based on Hofstede's cultural measurement theory to generate recommendations without the need to obtain personal data directly. The algorithm establishes links between users by analyzing their cultural characteristics, which helps to improve the accuracy of preference prediction. To further improve the results, a matrix factorization method is applied to identify hidden patterns in user preferences even in the absence of explicit system interaction data. The effectiveness of the approach proposed by the authors has been confirmed during experiments on the WS-Dream dataset. The results demonstrate that taking cultural factors into account can significantly improve the quality of recommendations, especially in cold-start environments. The integration of the matrix factorization method facilitates more accurate modeling of latent factors affecting user choice and allows recommendations to be adjusted according to the identified patterns. Incorporating cultural characteristics into the recommendation process outperforms conservative methods based solely on behavioral data and provides a more personalized approach to new users.