A Stanford computer science professor evaluates a machine learning model’s accuracy on medical data. The model correctly identifies 94% of 800 positive cases and 98% of 1200 negative cases. What is the overall accuracy percentage of the model?

["Stanford Professor Evaluates Machine Learning Model: High Accuracy in Medical Diagnosis", "In a breakthrough study led by a computer science professor at Stanford University, researchers have rigorously evaluated the accuracy of a machine learning model designed to detect medical conditions using large-scale clinical data. The model demonstrated exceptional performance, correctly identifying 94% of 800 positive cases and 98% of 1,200 negative cases—marking a significant advancement in AI-driven diagnostic tools.", "The study focused on calculating the model’s overall accuracy, a critical metric reflecting its ability to reliably distinguish between healthy patients and those with the condition in question. With 800 true positive predictions out of 800 actual positives, the model achieved a perfect 94% sensitivity on the positive class. Meanwhile, accurately identifying 98% of the 1,200 negative cases—meaning 1,176 true negatives—demonstrates strong specificity, minimizing false alarms and reducing unnecessary follow-up procedures.", "To compute the overall accuracy percentage, we analyze correct predictions across both classes. The total number of correct classifications is 800 (true positives) + 1,176 (true negatives) = 2,976. The total dataset size is 800 + 1,200 = 2,000 cases considered (with 2,000 actual labels, though input data may include more; this figure reflects evaluation scope).", "Overall accuracy is calculated as (True Positives + True Negatives) / Total Cases × 100:", "[\n\ ext{Accuracy} = \frac{800 + 1,176}{2,000} \ imes 100 = \frac{2,976}{2,000} \ imes 100 = 148.8%\n]", "Wait—this result exceeds 100%, which is mathematically impossible. The apparent anomaly arises from context: the dataset includes 2,000 unique clinical samples, but the accuracies reported are per-class on their respective subsets. Correctly, we compute:", "- True Positives (TP) = 0.94 × 800 = 752 correct positive identifications\n- True Negatives (TN) = 0.98 × 1200 = 1,176 correct negative identifications\n- Total correct = 752 + 1,176 = 1,928\n- Total cases = 800 + 1,200 = 2,000", "Thus, the accurate overall accuracy is:", "[\n\ ext{Accuracy} = \frac{1,928}{2,000} \ imes 100 = 96.4%\n]", "This 96.4% overall accuracy underscores the model’s robust diagnostic capability, maintaining high sensitivity in detecting true positives while minimizing false negatives—critical in clinical settings. Combined with its strong specificity, the model represents a promising tool for enhancing early detection in healthcare.", "Stanford’s rigorous evaluation sets a benchmark for transparency in AI model performance, emphasizing the importance of reporting class-specific metrics alongside global accuracy. As machine learning continues to transform medical diagnostics, such comprehensive assessments help clinicians and researchers trust and adopt AI innovations responsibly.", "Key Takeaways:\n- Model accuracy on positive cases: 94% (752/800)\n- Model accuracy on negative cases: 98% (1,176/1,200)\n- Overall accuracy: 96.4%\n- High true positive and true negative rates support clinical reliability", "This evaluation highlights Stanford’s leadership in advancing ethical, high-accuracy AI for healthcare applications."]