The prоcess whereby а rаdiоgrаphic image is created by variatiоns in absorption and transmission of the exiting x-ray beam is known as _______.
TBD2
An insurаnce cоmpаny is evаluating twо lоgistic regression models to flag new auto-insurance claims for a manual fraud-investigation review. A **false negative** means a fraudulent claim is paid out without review — the average fraudulent claim costs the company about **$7,200** in paid-out value plus recovery effort. A **false positive** means a legitimate claim gets slowed down for review — costing the customer a couple days of delay and the company about **$120** in investigator time and apology spend. Both models were evaluated on 3,000 recent claims. 180 of those claims were actually fraudulent. "Fraud" is the positive class. === MODEL F confusion matrix === - True Negatives (TN) = 2800 [actual: legit, predicted: legit] - False Positives (FP) = 20 [actual: legit, predicted: fraud] - False Negatives (FN) = 140 [actual: fraud, predicted: legit] - True Positives (TP) = 40 [actual: fraud, predicted: fraud] === MODEL G confusion matrix === - True Negatives (TN) = 2600 [actual: legit, predicted: legit] - False Positives (FP) = 220 [actual: legit, predicted: fraud] - False Negatives (FN) = 30 [actual: fraud, predicted: legit] - True Positives (TP) = 150 [actual: fraud, predicted: fraud] Answer the following: (1) Compute the accuracy, precision, and recall for each model (for the "Fraud" class). (2) Compute the total expected business cost of each model using the dollar figures above. (3) Which error type -- false positive or false negative -- is more harmful here, and which metric (precision or recall) should be prioritized? Justify. (4) Recommend a model. Cite specific numbers and acknowledge what the insurance company gives up by going with your recommendation.