Application of genetic algorithm and K-means algorithm for recommender systems

As there is an increasing emphasis on customer satisfaction, businesses are aiming to discover their customers' interests and preferences. Therefore, providing suitable products or services is crucial in today's competitive environment. Significant investments are being made in digitized services today. For example, the e-commerce sector has seen significant developments, taking advantage of the COVID-19 pandemic. One of these developments in e-commerce is the effective utilization of recommendation systems. Recommendation systems work by tracking interactions between users and items. Cluster analysis methods are used as a preprocessing step to reveal these interactions, grouping users or items with similar characteristics. Additionally, using clustering methods can help address some of the disadvantages in recommendation systems. In addressing these disadvantages, optimized clustering processes play a significant role. Optimized clustering processes can provide better initial recommendations for new users and items while addressing the data sparsity issue. The use of genetic algorithms in optimization is also quite popular. Based on the principles of natural selection and genetics, genetic algorithms are effective optimization algorithms used successfully in various fields and applications. In genetic algorithms, a starting population consisting of solution sequences is used, along with genetic operators like crossover and mutation. Using genetic algorithms to optimize clustering processes can have a positive impact on recommendation systems. This approach can lead to better clustering results in recommendation systems, reduced data sparsity, personalized recommendations, mitigating the cold start problem, and enhancing diversity. This approach aims to improve the performance of recommendation systems, ultimately enhancing user satisfaction and the effectiveness of recommendation results. In this thesis, we first examined traditional recommendation systems and their drawbacks. Then, clustering methods and their use in recommendation systems were detailed, followed by the introduction of genetic algorithms. Genetic algorithms were applied to optimize K-means clustering processes in the marketing dataset, and their performance was compared to K-means clustering. The aim of this study is to use the genetic algorithm clustering technique (GA-KOK) to better profile users and improve recommendation outputs, thereby enhancing the process of recommendation generation.