Data used for an object detection ML system was found to have been labelled incorrectly in many cases. Which ONE of the following options is most likely the reason for this problem? SELECT ONE OPTION

Incorrectly labeled data is a fundamental data quality issue. In supervised learning, the model learns to map inputs to outputs based on these labels (the "ground truth"). If the labels are wrong, the model learns incorrect patterns, leading to poor predictive performance. This is a direct cause of accuracy issues, as the model's ability to make correct predictions is compromised from the start.

A. Security issues: This relates to system vulnerabilities or malicious attacks, not the inherent correctness of the training data labels.
C. Privacy issues: This concerns the exposure of sensitive or personal information within the data, which is unrelated to label correctness.
D. Bias issues: While incorrect labeling could potentially introduce bias, the primary and most direct problem described is a lack of correctness, which is an accuracy issue.

1. ISTQB Certified Tester AI Testing (CT-AI) Syllabus

Version 1.0 (2021)

Section 2.2.2

Data Quality. The syllabus states

"The quality of data used for training

testing

and operating the AI-based system is a major factor for the quality of the system... Data quality has many characteristics

such as accuracy

completeness

consistency..."

2. Gorishniy

Y.

et al. (2021). "On the Importance of Data Quality for Machine Learning". arXiv preprint arXiv:2107.10472. Section 2 discusses how label noise (inaccurate labels) is a primary data quality problem that directly degrades model performance and accuracy.

3. Stanford University

CS229: Machine Learning

Course Notes. The course materials consistently emphasize that the quality and correctness of labeled data are prerequisites for training accurate supervised learning models.