Multivariate CNN-LSTM Model for Multiple Parallel Financial Time-Series Prediction

Table 4

Comparison of descriptive statistics between actual and predicted time-series data produced by the proposed multivariate CNN-LSTM model in testing period, i.e., last 72 trading days of 2020. Note: Act = actual and Pred = predictions.

Descriptive statistics	Hang Seng (HSI)		Japan (N225)		Singapore (STI)		Indonesia (JSX)
	Act	Pred	Act	Pred	Act	Pred	Act	Pred
Mean	1.1917	1.1802	1.0997	1.0853	1.0260	1.0221	1.0315	1.0421
Median	1.1872	1.1783	1.0956	1.0847	1.0336	1.0401	1.0093	1.0102
Standard deviation	0.0320	0.0296	0.0677	0.0643	0.0277	0.0291	0.0787	0.0766
Sample variance	0.0010	0.0012	0.0046	0.0043	0.0008	0.0071	0.0062	0.0058
Kurtosis	0.3295	0.3187	1.7725	1.7523	0.8161	0.8021	1.3399	1.3187
Skewness	0.3376	0.3152	0.1284	0.1179	1.1910	1.1876	0.3793	0.3685
Range	0.1497	0.1502	0.2022	0.2013	0.1132	0.1045	0.2520	0.2471
Minimum	1.1288	1.1256	1.0121	1.0097	0.9570	0.9613	0.9226	0.9197
Maximum	1.2786	1.2691	1.2143	1.2095	1.0702	1.0273	1.1746	1.1637
Confidence level (95.0%)	0.0075	0.0086	0.0159	0.0137	0.0065	0.0061	0.0185	0.0157