<![CDATA[This study aims to evaluate and compare the performance of two recommendation system approaches, namely Collaborative Filtering (CF) and Content-Based Filtering (CBF), in providing relevant product recommendations to users in an e-commerce context. The dataset used consists of 120 data including 90 relevant and recommended products (True Positive), 20 recommended but irrelevant products (False Positive), and 10 relevant but not recommended products (False Negative). Based on the calculation results, both methods show a precision value of 0.818 and a recall of 0.900. This means that approximately 81.8% of products recommended by the system are truly relevant, while 90% of the total relevant products are successfully recommended to users. The F1-score value obtained of 0.857 illustrates a good balance between the accuracy and completeness of the recommendations generated by the system. Furthermore, to measure the level of rating prediction error, the Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) metrics are used. The evaluation results show that the CF method has an MSE value of 0.0784 and an RMSE of 0.28, while the CBF method shows an MSE of 0.0961 and an RMSE of 0.31. The lower RMSE value of CF indicates that this method has better accuracy in predicting user preferences than CBF. Overall, both methods show good performance with a low error rate. However, CF proved slightly superior in providing recommendations that match user preferences, so it can be used as a basis for developing smarter and more personalized recommendation systems on e-commerce platforms.]]>
