Question

a manufacturing company wants to detect defects in its products using images from its production line. they have thousands of product images, but havent labeled them as defective or non - defective yet.
broadly supervised learning
unsupervised learning to detect anomalies
skip machine learning and use rule - based image processing

a financial institution wants to automate its loan approval process. they have extensive historical data on past loans, including whether they were repaid or defaulted.
supervised learning with a decision tree
unsupervised learning with clustering
no machine learning needed - use credit score only

Explanation:

1. For the manufacturing company, since the product - images are unlabeled, supervised learning cannot be used directly as it requires labeled data. Unsupervised learning for anomaly detection can be a good approach to find patterns that may indicate defects. Rule - based image processing might be too rigid without prior knowledge of defect characteristics.
2. For the financial institution, with extensive historical data on past loans (including repayment or default status), supervised learning with a decision tree can be used to learn from the existing labeled data (repayment or default) to predict whether new loan applications should be approved or not. Unsupervised learning with clustering is not appropriate as it does not use the labeled information of repayment or default. Using only credit scores may not fully utilize all the available historical data.

Answer:

1. Unsupervised learning to detect anomalies
2. Supervised learning with a decision tree