Predictive Analytics of Student Performance

Data Loading and Exploration

• Import CSV datasets for Math and Portuguese performance, ensuring the correct delimiter settings.

• Examine the summary of both datasets and contrast them with the metadata. Identify at least two notable distributions across columns.

Predicting Final Grade (Linear Regression)

• Build linear regression models to predict the final grade (G3) based on one or more previous grades (G1, G2).

• Split data into 70% training and 30% testing.

• Evaluate the model’s goodness-of-fit (R²) and visualize the relationship between actual and predicted values using scatter plots.

• Compare models with different predictor variables (G1, G2, G1+G2, all variables) to assess which model is most accurate and useful.

Predicting Risk of Failing (Logistic Regression)

• Create a new column to classify students as passing or failing based on their average grades.

• Build a logistic regression model to predict “PassFail” using all other variables.

• Split data into training and testing sets, and visualize the model’s performance with a confusion matrix.

• Identify any issues with the model and propose improvements based on the confusion matrix.

Decision Tree for Risk Prediction

• Build a decision tree model using the “PassFail” column and excluding grade variables (G1, G2, G3).

• Analyze the decision tree to identify important variables and patterns or rules that predict student outcomes.

• Evaluate the model’s utility in identifying students at risk and reflect on whether it should be used for interventions.

Random Forest for Risk Prediction

• Apply a random forest model using the same dataset and “PassFail” column.

• Adjust threshold values and assess the model’s performance through confusion matrices.

• Compare the random forest’s effectiveness with other models and determine which threshold is optimal for identifying students at risk.