Predicting Novels Popularity Based on Optimized Random Forest using Success History Intelligent Optimization

Reading novels help people to develop their social skills and mainly helps to reduce stress. In that, predicting novel popularity becomes a difficult task that is impacted by amount of variables, including reader preferences, sales patterns, and literary qualities. Machine Learning (ML) techniques are highly recommended for prediction task wherein using the Random Forest (RF), the popularity of sentiment as well as emotional variables derived from the narrative text of novels has been predicted. Since these techniques are heavily relied on hyperparameter values because, when optimized hyperparameters have been defined and modified during the training process, their predicted accuracy could improve. In order to fine-tune the RF hyperparameters, a Success History Intelligent Optimisation (SHIO) is presented for the prediction of popularity of online novels. A total of 2,439 eBooks/novels around nine distinct categories in the online publishing from Project Gutenberg. By addressing the issue of consumer information overload, these vectors could become beneficial for advancing the development of novel recommender systems. The SHIO-RF achieves a lower Mean Absolute Error (MAE) of 0.15 and outperforms the compared conventional algorithms including Shapley Additive explanations and Light Gradient Boosting Machine Classifier (SHAP-LGBM), Attention Mechanism based Graph Neural Network (AM-GNN) and RF.