Model registry in ML Python - Model Metrics & Evaluation

Metrics & Evaluation - Model registry

Which metric matters for Model Registry and WHY

A model registry is a system to store and manage machine learning models. It tracks versions, metadata, and performance metrics. The key metrics here are model performance metrics like accuracy, precision, recall, and F1 score. These metrics help decide which model version is best to use or deploy.

Why? Because the registry helps compare models fairly and pick the best one. It also stores metadata like training data, parameters, and evaluation results to keep track of model quality over time.

Confusion Matrix or Equivalent Visualization

When models are registered, their confusion matrices are often saved as part of their evaluation. Here is an example confusion matrix for a binary classifier:

      | Predicted Positive | Predicted Negative |
      |--------------------|--------------------|
      | True Positive (TP) = 80  | False Positive (FP) = 10 |
      | False Negative (FN) = 20 | True Negative (TN) = 90  |
    

This matrix helps calculate precision, recall, and accuracy, which are stored in the registry for each model version.

Precision vs Recall Tradeoff with Examples

Models in the registry often have different precision and recall values. Choosing which model to deploy depends on the tradeoff:

• High Precision: Few false alarms. Good for spam filters so real emails are not marked spam.
• High Recall: Few missed cases. Important for cancer detection to catch all patients.

The registry helps compare these tradeoffs by storing both metrics for each model version.

What Good vs Bad Metric Values Look Like

Good model metrics in the registry mean:

• High accuracy (e.g., > 90%) on test data
• Balanced precision and recall (e.g., both > 80%)
• Stable performance across versions

Bad metrics mean:

• Low accuracy (e.g., < 60%)
• Very low recall or precision (e.g., < 50%)
• Big drops in performance compared to previous versions

The registry flags these to avoid deploying poor models.

Common Metrics Pitfalls in Model Registry

• Accuracy Paradox: High accuracy but poor recall on rare classes can mislead model choice.
• Data Leakage: Metrics look great if test data leaks into training, but real-world performance drops.
• Overfitting: Metrics very high on training but low on test data indicate overfitting.
• Ignoring Metric Context: Not considering which metric matters for the problem leads to wrong model selection.

Self Check: Is a Model with 98% Accuracy but 12% Recall on Fraud Good?

No, this model is not good for fraud detection. Even though accuracy is high, recall is very low. This means it misses most fraud cases, which is dangerous. In fraud detection, catching fraud (high recall) is more important than just overall accuracy.

The model registry would flag this model as unsuitable for deployment despite the high accuracy.

Key Result

Model registry tracks key metrics like precision, recall, and accuracy to help pick the best model version for deployment.

Practice

1.

What is the main purpose of a model registry in machine learning?

easy

A. To store and organize machine learning models

B. To train machine learning models faster

C. To clean and prepare data for training

D. To visualize model predictions

Solution

1. Step 1: Understand the role of a model registry

   A model registry is designed to keep track of machine learning models, their versions, and metadata.

2. Step 2: Compare options with the definition

   Only To store and organize machine learning models describes storing and organizing models, which matches the main purpose of a model registry.

3. Final Answer:

   To store and organize machine learning models -> Option A

4. Quick Check:

   Model registry = store and organize models [OK]

Hint: Model registry = place to keep models safe and sorted [OK]

Common Mistakes:

• Confusing model registry with data cleaning
• Thinking it speeds up training
• Mixing it with visualization tools

2.

Which of the following is the correct way to add a model to a simple Python dictionary used as a model registry?

model_registry = {}
# Add model here

easy

A. model_registry['model_v1'] = trained_model

B. model_registry.add('model_v1', trained_model)

C. model_registry.append('model_v1', trained_model)

D. model_registry.insert('model_v1', trained_model)

Solution

1. Step 1: Recall how to add key-value pairs to a dictionary

   In Python, adding an item to a dictionary uses the syntax dict[key] = value.

2. Step 2: Check each option for correct dictionary syntax

   The incorrect options use methods like .add(), .append(), and .insert() which are not available for dictionaries. Only model_registry['model_v1'] = trained_model uses the correct assignment syntax.

3. Final Answer:

   model_registry['model_v1'] = trained_model -> Option A

4. Quick Check:

   Dictionary add = dict[key] = value [OK]

Hint: Use dict[key] = value to add items in Python [OK]

Common Mistakes:

• Using list methods like append on dictionaries
• Trying to call add or insert on dict
• Confusing dictionary with list syntax

3.

Given the following code, what will be the output?

model_registry = {}
model_registry['v1'] = {'accuracy': 0.8}
model_registry['v2'] = {'accuracy': 0.85}
print(model_registry['v2']['accuracy'])

medium

A. 0.8

B. 0.85

C. {'accuracy': 0.85}

D. KeyError

Solution

1. Step 1: Understand the dictionary structure

   model_registry has keys 'v1' and 'v2', each storing a dictionary with an 'accuracy' key.

2. Step 2: Access the nested value

   model_registry['v2']['accuracy'] accesses the accuracy value for version 'v2', which is 0.85.

3. Final Answer:

   0.85 -> Option B

4. Quick Check:

   Access nested dict value = 0.85 [OK]

Hint: Access nested dict with dict[key1][key2] [OK]

Common Mistakes:

• Confusing keys and values
• Expecting full dict instead of value
• Mistyping keys causing KeyError

4.

Identify the error in this code snippet that tries to update a model's accuracy in a registry:

model_registry = {'v1': {'accuracy': 0.8}}
model_registry.update('v1', {'accuracy': 0.85})

medium

A. Syntax error due to missing colon

B. update cannot update nested dictionaries

C. The key 'v1' does not exist in the dictionary

D. update method expects a dictionary, not two arguments

Solution

1. Step 1: Check the usage of dictionary update method

   The update() method takes a single dictionary or iterable of key-value pairs, not separate arguments.

2. Step 2: Identify the correct way to update nested dictionary

   To update nested dict, access it first: model_registry['v1'].update({'accuracy': 0.85}).

3. Final Answer:

   update method expects a dictionary, not two arguments -> Option D

4. Quick Check:

   dict.update() needs one dict argument [OK]

Hint: dict.update() takes one dict, not multiple args [OK]

Common Mistakes:

• Passing multiple arguments to update()
• Trying to update nested dict without accessing it
• Assuming update() works like assignment

5.

You want to keep track of multiple versions of a model with their accuracy and deployment status. Which data structure best suits a simple model registry in Python?

hard

A. A set of model names without details

B. A list of tuples with version and accuracy only

C. A dictionary with version keys and nested dictionaries for metrics and status

D. A single string describing the latest model

Solution

1. Step 1: Identify the need for storing multiple details per version

   You need to store version keys, accuracy, and deployment status, which requires nested information.

2. Step 2: Evaluate data structures for this need

   A dictionary with version keys and nested dictionaries for details fits best. Lists or sets lack easy key-based access or nested info.

3. Final Answer:

   A dictionary with version keys and nested dictionaries for metrics and status -> Option C

4. Quick Check:

   Nested dicts store complex model info best [OK]

Hint: Use nested dicts for versions and details [OK]

Common Mistakes:

• Using lists or sets that lack keys
• Storing only one detail per model
• Using strings that can't hold structured data