Users’ Rating Predictions Using Collaborating Filtering Based on Users and Items Similarity Measures

Table 8

Comparison of Applied Algorithms on Datasets as a whole.

Algorithms	RMSE			MAE
	ML 100k	ML 1M	Ciao DVD	ML	ML 1M	Ciao DVD
				100k
User K-NN (cosine similarity)	0.947	0.899	0.975	0.746	0.706	0.737
User K-NN (Pearson correlation)	0.939	0.887	0.969	0.739	0.696	0.737
Item K-NN (cosine similarity)	0.934	0.883	0.975	0.736	0.694	0.736
Item K-NN (Pearson similarity)	0.933	0.879	0.964	0.734	0.690	0.734
Slope one	0.951	0.902	1.093	0.749	0.711	0.816
Random	2.105	2.138	2.307	1.723	1.762	1.888
Global average	1.131	1.115	1.079	0.950	0.933	0.833
User item baseline	0.952	0.909	0.976	0.755	0.720	0.762
Matrix factorization	0.988	0.954	1.107	0.771	0.746	0.860
Biased MF	0.992	0.964	1.032	0.770	0.752	0.797
Factor wise MF	0.989	0.940	1.622	0.765	0.731	1.215