A prediction model that had been trained and optimize for multi-label classification for three different classes of quality of fruits and three different classes of types of fruits were tested with test dataset distribution as shown in Table Q4(a). The result of the classification is shown in the form of a Confusion matrix as seen in Table Q4(b). (a) The confusion matrix given in Table Q4(b) is incomplete. Calculate the value of J1, J2, J3 and J4 respectively. [8 marks) (b) Reconstruct the multi-label confusion matric into multi-class confusion matric for different types of fruit classification. Assume that the testing dataset in Table Q4(a) is tested with the same trained network. [10 marks] C) (c) Based on the confusion matrix answered in Q4(b), for eggplant classification: (1) Calculate the True-Positive (TP), True-Negative (TN), False-Positive (FP) and False-Negative (FN) [2 marks] (ii) Find the Fl-score and specificity of the classification (5 marks) Table Q4(a) No of test samples Class Chilli=c Type of Fruits Eggplant- Tomato =T E 110 115 125 110 95 125 Quality of Fruits Grade 'A' = A Unrippen = U Infested = 1 100 120 115 No of test samples Table Q4(b) ACTUAL AE UE IE User's accuracy AC UC IC AT UT IT AC 100 3 0 0 3 0 0 90% UC 0 10000 0 0 0 0 100% IC 0 J1 115 J2 0 0 0 0 93% 0 | 0 | o o 0 0 AE 0 N 0 95 0 3 0 5 P R E D I I 110% UE 0 0 120 0 0 5 95% 0 1 0 J3 0 5 IE 0 5 0 95 6 0 5 J4 OZ N AT 0 2 0 0 0 0 100 0 5 93% UT 0 1 0 5 0 0 3 110 5 89% IT 0 2 0 0 0 0 0 0 100 98% Producer's accuracy 100% 83% 100% 86% 96% 100% 87% 100% 80%