Success in books: predicting book sales before publication

EPJ Data Science

Table 1 Robustness experiment. To analyze the robustness of our model assuming a realistic setting, we trained our model using earlier instances and tested it on books published later. We report AUC (top) and High-end RMSE (bottom) scores for models train on qth quarter of the year 2015 and test on the \(q+1\)th quarter. We observe that Learning to Place almost always outperform other methods

Quarter	Quarter 2	Quarter 3	Quarter 4	Quarter 2	Quarter 3	Quarter 4
Category	Fiction (AUC)			Nonfiction (AUC)
KNN	0.83	0.82	0.82	0.81	0.80	0.82
Linear Regression	0.85	0.83	0.86	0.85	0.84	0.86
Neural Network	0.88	0.83	0.73	0.83	0.83	0.85
Learning to Place	0.88	0.85	0.88	0.85	0.83	0.85
Category	Fiction (High-end RMSE)			Nonfiction (High-end RMSE)
KNN	0.60	0.91	1.03	0.71	0.72	0.77
Linear Regression	0.61	0.77	0.89	0.58	0.61	0.58
Neural Network	0.44	0.81	2.83	0.71	0.57	0.63
Learning to Place	0.42	0.71	0.45	0.49	0.51	0.62