Technical Name Is Ground Truth Always Correct? An Explainable Analytics Ecosystem: Understanding Misjudgments for Enhancing Decision-makings in MedicineRecommender Systems
Project Operator National Yunlin University of ScienceTechnology
Project Host 白浩廷
Summary
TC can identify all positive (e.g., malignant) observations at low ratios of training to testing data, e.g., 1:9 using the Breast Cancer Wisconsin (Original) dataset. Without fine-tuned parametersrandom selection, the uncertainty of the methodology is eliminated when using TC. TC would be useful for reducing misjudgments (e.g., diagnostic errors)avoiding waste of medical resourceslegal disputes. On the other hand, a TC-based recommender system can achieve around 50 hit rate of successful recommendation in the online retail dataset. The most import point is that TC do not need profile of customers,therefore, TC is adequate for the EU’s AI regulation: fairnessprivacy.
Scientific Breakthrough
"1. Without fine-tuned parametersrandom selection, the uncertainty of the methodology is eliminated when using TC. 
 
 2. TC can visualize causes,prediction errors in the network are traceablecan be corrected. 
 
 3. TC can identify all positive (e.g., malignant) observations at low ratios of training to testing data, e.g., 1:9 using the Breast Cancer Wisconsin (Original) dataset. 
 
 4. TC shows potential in identifying whether the ground truth is incorrect (e.g., diagnostic errors).
 
 5. The TC-based recommender system can achieve around 50 hit rate of successful recommendation in the online retail dataset. TC do not need profile of customers,therefore, TC is adequate for the EU’s AI regulation: fairnessprivacy.
Industrial Applicability
The ground truth is not always correct, e.g., diagnostic errors globally it may result in 7 million children died yearly,in US the rate is 5.08 that approximates to 12 million adults yearly. TC possesses explainabilityreproducibility, which is useful for discovering the reason behind failure. In medical diagnosis, TC can be the second opinion for reducing diagnostic errors. On the other hands, recommender systems have been widely applied to many areas such as ecommerce “Tonight, I’ll be eating.”, retail “You may also like”,so on. TC-based recommender system can achieve high hit rate of successful recommendation within the AI regulations of EU such as fairnessprivacy. It would have great potential in e-commerce.
Keyword Explainable AI (XAI) Machine Learning Anomaly Detection Classification Recommender Systems Medical Diagnosis Diagnostic Errors AI Regulations Reproducibility Privacy
Notes
  • Contact
other people also saw