Comparative Analyses of Response Surface Methodology and Artificial Neural Network on Medium Optimization for Tetraselmis sp. FTC209 Grown under Mixotrophic Condition
Table 5
Predicted lipid productivity by RSM and ANN together with the residual error functions (, RMSE, and MAE).
Run
RSM predicted lipid productivity
RSM absolute deviation
ANN predicted lipid productivity
ANN absolute deviation
1
37.150
5.430
31.715
0.005
2
56.735
2.595
59.333
0.003
3
53.952
4.848
58.750
0.050
4
60.632
2.088
62.719
0.001
5
57.594
2.736
60.333
0.003
6
61.140
0.030
61.473
0.303
7
61.356
2.224
63.583
0.003
8
53.997
8.263
62.356
0.096
9
74.311
0.251
74.063
0.002
10
82.909
6.509
74.961
1.439
11
58.749
2.371
61.760
0.640
12
65.189
9.189
54.854
1.146
13
39.258
0.078
35.486
3.694
14
49.187
6.747
43.021
0.581
15
169.466
6.696
171.234
8.664
16
169.466
3.954
171.234
1.986
17
169.466
7.914
171.234
5.946
18
169.466
11.156
171.234
13.124
19
169.466
7.154
171.234
5.186
RSM Model = 0.985, ANN testing set = 0.993. RSM Model RMSE = 5.719, ANN testing set RMSE = 4.176. RSM Model MAE = 4.750, ANN testing set MAE = 2.256.
