IB Computer Science A.4.8 Predictive Modeling in Data Mining Notes

Key Techniques in Predictive Modeling

1. Decision Tree Induction

1. Decision trees are a popular method for predictive modeling due to their simplicity and interpretability.
2. How Decision Trees Work:
   1. Splitting the Data: The tree starts with the entire dataset and splits it into subsets based on the value of an attribute.
   2. Recursive Partitioning: This process is repeated recursively for each subset, creating branches until a stopping criterion is met (e.g., all data in a subset belong to the same class).
   3. Prediction: To make a prediction, the model traverses the tree from the root to a leaf node, following the path determined by the input attributes.

2. Backpropagation in Neural Networks

1. Neural networks are powerful models inspired by the human brain, capable of capturing complex patterns in data.
2. How Backpropagation Works:
   1. Forward Pass: The input data is passed through the network, and predictions are made.
   2. Error Calculation: The difference between the predicted and actual values (the error) is calculated.
   3. Backward Pass: The error is propagated backward through the network, and the weights are adjusted to reduce the error.
   4. Iteration: This process is repeated for many iterations until the model achieves satisfactory accuracy.

3. Row Selection for Predictions

1. Predictive modeling often involves identifying which rows (records) in a database are most useful for making accurate predictions.
2. Key Steps:
   1. Feature Selection: Identifying the most relevant attributes (features) that influence the outcome.
   2. Sampling: Selecting a representative subset of the data for training the model.
   3. Validation: Using a separate subset of data to test the model's accuracy.