An Empirical Study on Downstream Dependency Package Groups in Software Packaging Ecosystems

Table 11

Prediction results of number of releases of CDDGs.

Prediction models		CDDG size  100			CDDG size  100
		Cargo	CPAN	RubyGems	Cargo	CPAN	RubyGems
Linear regression	MAE	3.57	1.52	6.56	3.72	6.87	11.02
	RMSE	5.08	2.46	7.89	5.06	8.63	17.06
	MAPE	18.07%	11.26%	23.82%	26.95%	36.85%	36.50%
Random forest	MAE	2.51	1.43	5.74	3.16	6.57	9.70
	RMSE	2.90	2.27	7.36	4.45	8.45	14.07
	MAPE	16.10%	9.59%	21.75%	20.30%	30.46%	34.31%
KNN	MAE	2.54	1.41	7.24	3.74	8.02	9.72
	RMSE	2.82	2.13	8.55	5.10	9.84	13.57
	MAPE	15.70%	9.12%	26.20%	24.15%	36.00%	33.03%
AdaBoost	MAE	3.23	1.79	7.18	3.80	7.92	12.94
	RMSE	3.92	2.58	9.30	4.97	9.12	15.49
	MAPE	21.12%	12.14%	24.56%	22.14%	37.72%	37.71%
GBRT	MAE	2.70	1.52	6.18	3.59	6.42	9.62
	RMSE	3.37	2.43	7.70	5.00	9.03	14.21
	MAPE	18.70%	10.43%	22.92%	22.58%	28.49%	36.76%

Highlight the optimal results under each indicator of different models.