Word Sequential Using Deep LSTM and Matrix Factorization to Handle Rating Sparse Data for E-Commerce Recommender System

Table 7

Performance comparison of LSTM-PMF over the state-of-the-art methods on ML-10M.

Sparseness level (high-low)	RMSE evaluation result			Comparison result
	PMF	CNN-PMF	LSTM-PMF	PMF versus LSTM-PMF (%)	CNN-PMF versus LSTM-PMF (%)
10% (90% sparseness level)	1.27539	0.93629	0.95506	25.10	−2
20% (80% sparseness level)	1.05233	0.89332	0.89117	15.30	0.24
30% (70% sparseness level)	0.96513	0.86621	0.85185	11.70	1.65
40% (60% sparseness level)	0.91827	0.84673	0.82737	9.89	2.28
50% (50% sparseness level)	0.88834	0.83604	0.81567	8.18	2.43
60% (40% sparseness level)	0.86673	0.82794	0.80968	6.58	2.20
70% (30% sparseness level)	0.85071	0.82054	0.80276	5.63	2.16
80% (20% sparseness level)	0.84049	0.81276	0.79735	5.13	1.89
90% (10% sparseness level)	0.82796	0.80505	0.7902	4.56	1.84
∑ (total)				92	13
(average)				10.23	1.44