<![CDATA[This research is motivated by the difficulty viewers have in finding movies that suit their tastes amid the large number of movies being produced. Current movie ratings are often based solely on direct assessments by viewers without considering factors such as genre, audience age category, and movie synopsis. This study aims to predict movie ratings using the Neural Factorization Machines (NFM) approach. The research method includes data preparation, which covers dataset file merging, age category mapping, data cleaning, text conversion to lowercase, regular expression removal, removal of non-English text, tokenization, lemmatizing, word embedding, one-hot encoding, and label encoding. The modeling process was carried out by building an NFM model consisting of feature inputs, embedding layers, bi-interaction layers, hidden layers, and prediction scores. Model evaluation was carried out by setting hyperparameters, namely epoch and batch size, to optimize model performance. This study was conducted with 9 tests using a combination of epochs (30, 50, and 100) and batch sizes (64, 128, and 256). The evaluation results show that the lowest MSE value, which means the best, in the training data is 1.181 with a batch size of 256 and an epoch of 100, and in the validation data is 1.230 with a batch size of 256 and an epoch of 100. However, in the test data, the configuration with a batch size of 128 and 50 epochs gave the best MSE of 1.280. Although the model showed the best performance in the training and validation data with a batch size of 256 and 100 epochs, the evaluation graph indicated overfitting. These findings show that the NFM model is capable of predicting movie ratings based on genre, audience age category, and movie plot description.]]>
