Imbalanced Datasets

Requirements Clarification & Assessment

1. Understanding the Imbalance:

   • Assess the degree of imbalance in the dataset by calculating the ratio of the majority class to the minority class.
   • Determine if the imbalance is severe, moderate, or mild, as this will influence the choice of strategy.

2. Defining the Objective:

   • Clarify the primary goal of the model. Is it to maximize accuracy, precision, recall, or a combination of these metrics?
   • Identify the business impact of misclassifying the minority class versus the majority class.

3. Data Characteristics:

   • Examine the dataset for size, feature distribution, and potential noise.
   • Identify if additional data collection is feasible to address the imbalance.

4. Evaluation Metrics:

   • Decide on appropriate evaluation metrics (e.g., Precision, Recall, F1 Score, Matthews Correlation Coefficient) that reflect the importance of correctly classifying the minority class.