<![CDATA[Information about games will continue to grow as technology develops. Information that is too fast to develop and be accepted by ordinary people then becomes overloaded. One solution to this problem is a recommendation system. However, due to the ever-increasing information, the recommendation system process time is longer. Classification can help the recommendation system in grouping the game information into various categories.Therefore, the purpose of this study is to compare the recommendation results and processing time of a game recommendation system with classification and without classification. The data used is game data from the STEAM platform. The recommendation system method used is content-based filtering by comparing a keyword with a game description. For its classification, it uses multilabel classification using the LSTM model. The games will be classified according to their genre using the LSTM model. As a result, the unclassified content-based filtering recommendation system resulted in an average precision of 60%, and an average processing time of 9.72 seconds. Then in the classification, the best model LSTM was obtained with an accuracy of 50.7%. With the obtained model, it is then used in a Content-Based Filtering recommendation system that produces an average precision of 55%, and an average processing time of 2 seconds.Information about games will continue to grow as technology develops. Information that is too fast to develop and be accepted by ordinary people then becomes overloaded. One solution to this problem is a recommendation system. However, due to the ever-increasing information, the recommendation system process time is longer. Classification can help the recommendation system in grouping the game information into various categories.Therefore, the purpose of this study is to compare the recommendation results and processing time of a game recommendation system with classification and without classification. The data used is game data from the STEAM platform. The recommendation system method used is content-based filtering by comparing a keyword with a game description. For its classification, it uses multilabel classification using the LSTM model. The games will be classified according to their genre using the LSTM model. As a result, the unclassified content-based filtering recommendation system resulted in an average precision of 60%, and an average processing time of 9.72 seconds. Then in the classification, the best model LSTM was obtained with an accuracy of 50.7%. With the obtained model, it is then used in a Content-Based Filtering recommendation system that produces an average precision of 55%, and an average processing time of 2 seconds.]]>
