University:
University of South Wales
Course:
G40C | Computer Science
Academic year:

2023
Views:

105

Pages:

9
Author:

Rebecca L

MODULE-3 Classifier Performance K-cross field validation 1. Divide the sample data into k parts 2. Use k-1 parts for training and 1 for testing 3. Repeat the procedure 'k' times. 4. Determine the expected performance based on the classified result Boot Strapping straining * Bootstrapping is a validation method in which dataset strength should be always fired and once the testing data removed from the dataset the positions of the removed data cam be occupied by the replicated training data Bootstrap2: Bootstrap3: 22 23 23 24 25 26 5x2 Validation It is a 5 time 2 fold cross validation. Training Testing ② * Here, training data and testing data are exchanged during each iteration and the procedure should repeat fou 5 times. desting Training x_{1} x_{2} are fold of data haining Bet Validation Set T_{t} = X_{2} T_{11} = x_{1} Upsilon_{T} = X_{2} T_{t} = X_{1} Upsilon_{4} = x_{4} ② Tr = x_{3} tau_{tau} = x_{4} T_{k} = x_{3} tau_{n} = x_{5} tau_{r} = x_{6} T_{t} = x_{6} T_{t} = x_{5} tau_{n} = 7x_{1} Tr =X8 T_{r} = X_{q} S_{r} = X_{10} T_{t} = X_{q} T_{t} = X_{t} T_{t} = X_{10} T_{t} = x_{9} Classification Errors * Outcomes of classification are: cis fave Positive (ii False Pasitive (ii) Chue Negative (iv) False Negative True Positive *Let the true class label of x be c, if the model predicts the class field labels a forc, than that sample is called Taue Positive False Positive * het the true class dabel of a be c, if the model predicts the class de as TC, then that sample is False positive. True Negative *het the true class label of a be tc, if the model predicts the class label as C, then that sample is called true Negative. False Negative het the true class label of & be tre, if the model predicts the class label as ic, then that sample is called False Negative. Confusion Matrix The matrix representation of force positive, False positive, True Negative and False Negative is called Confusion Matrix Actual class label c Actual class label C Predicted class label C Tave positive false Positive Predicted class label False Negative chance Negative Precision and Recall Precision (p) = Taue Positive True Positive + False Positive = TP TP+FP Recall = True Positive True Positive + False Negative TP TP+FN Example: Suppose a computer program for recognizing dogs in a photograph identifies & dogs in a picture containing 12 dogs and some cats. of the 8 8 dogs identified, 5 actually are dogs while the rest are cats. Compute precision and recall. ⇒ Precision = 5/8 5/6+3] Recall = 5/12 (5/115] fane Positive Rate TP TP+FN for action of positive examples correctly classified.