SUL Topic 3 - Logistic Regression

Question 1

Logistic Regression

Answer 1

A statistical method used to predict binary outcomes by analyzing the outcome’s relationship with one or more predictor variables.

Question 2

1) Obtaining historical data

Answer 2

The first step in building a logistic regression model is to obtain historical data with a labeled column for the outcome you want to predict.

Question 3

2) Partitioning data

Answer 3

Divide the data into a training set and a testing set, with the training set being larger, to build and test the model respectively.

Question 4

3) Selecting relevant variables

Answer 4

Select relevant variables from the training data based on logic, domain expertise, or theory to build the model.

Question 5

4) Building the model

Answer 5

The logistic regression algorithm creates a model, which is an equation, based on the selected variables and outcome.

Question 6

5) Testing model accuracy

Answer 6

Test the accuracy of the model using the test set and a confusion matrix.

Question 7

Training and test set division

Answer 7

Divide the data randomly into a training set (2/3) and a test set (1/3), unless it is a time series problem.

Question 8

Cross-validation

Answer 8

Divide the data into k subsamples, use k-1 for training and the remaining one for testing, repeat k times to ensure all examples are used for both training and testing.

Question 9

Data balancing

Answer 9

Balance the training data by deleting records from the most frequent category or duplicating records in the less frequent category to avoid poor models.

Question 10

1st Consideration:
Overfitting

Answer 10

When a model is excessively complex and includes noise instead of underlying relationships, which can be avoided through cross-validation.

Question 11

2nd Consideration:
Variable selection challenges

Answer 11

Deciding on the input variables to use can be challenging and is often based on logic, domain expertise, or theory.

Question 12

3rd Consideration:
Minimum data requirements

Answer 12

A minimum of 50 cases per predictor is recommended, and larger datasets with balanced categories tend to produce better results.

Question 13

Model performance evaluation

Answer 13

• Evaluate model performance using
  ROC curve
• Misclassification rate
• Accuracy rate
• Confusion matrix.

Question 14

Improving model performance

Answer 14

• Adjust model parameters
• Try different modeling techniques
• Improve data quality
• Experiment with multiple models.

Question 15

Selecting target and input variables (Summary #1)

Answer 15

Select a binary target variable to predict and relevant input variables with few missing values and low correlation.

Question 16

Preparing data (Summary #2)

Answer 16

Deal with missing, dirty, or duplicate data, remove outliers, and split data into training and validation sets or use cross-validation.

Question 17

Building and running the model (Summary #3)

Answer 17

Observe the statistical significance of each predictor variable and use a confusion matrix to compare and select the best model.

Question 18

Model validation (Summary #4)

Answer 18

Apply the model to the validation set to assess its accuracy on new data and avoid overfitting.