Table 1:

Each row shows the average absolute error for a different combination of model components. B^G signifies the overall bias, B^U and B^M signify the user and message biases. F^U and F^M indicate the user and message features. M F indicates the matrix factorization model with K = 1.

Model	Question 1	Question 2	Question 3	Question 4
BG	0.887 ± 0.029	0.902 ± 0.029	0.773 ± 0.030	0.892 ± 0.028
BG, FU	0.902 ± 0.031	0.928 ± 0.030	0.855 ± 0.032	0.930 ± 0.031
BG, FM	0.884 ± 0.029	0.890 ± 0.029	0.770 ± 0.029	0.878 ± 0.028
BG, BU	0.729 ± 0.029	0.750 ± 0.029	0.723 ± 0.029	0.736 ± 0.030
BG, BM	0.870 ± 0.029	0.855 ± 0.030	0.794 ± 0.028	0.822 ± 0.031
BG, M F	0.798 ± 0.031	0.800 ± 0.029	0.792 ± 0.030	0.841 ± 0.032
BG, FU, FM	0.906 ± 0.030	0.925 ± 0.030	0.860 ± 0.033	0.924 ± 0.032
BG, BU, BM	0.714 ± 0.028	0.734 ± 0.028	0.715 ± 0.027	0.731 ± 0.028
BG, FU, M F	0.794 ± 0.031	0.805 ± 0.028	0.865 ± 0.032	0.930 ± 0.034
BG, FM, M F	0.792 ± 0.030	0.776 ± 0.028	0.774 ± 0.028	0.847 ± 0.031
BG, BU, M F	0.775 ± 0.031	0.759 ± 0.028	0.724 ± 0.027	0.724 ± 0.029
BG, BM, MF	0.775 ± 0.028	0.754 ± 0.028	0.766 ± 0.028	0.800 ± 0.030
BG, FU, FM, M F	0.810 ± 0.031	0.816 ± 0.031	0.870 ± 0.039	0.898 ± 0.033
BG, BU, BM, M F	0.750 ± 0.029	0.732 ± 0.027	0.697 ± 0.027	0.720 ± 0.028