The document analyzes emails from the Enron bankruptcy scandal using predictive analytics. It cleans and pre-processes a dataset of 158 users' emails, creates bags of words and extracts frequent words. It then classifies the sentiments and applies a CART model to train and test datasets, checking the accuracy with confusion matrices, ROC curves and AUC scores. The CART model depicts key terms like "California" and "Gas" and achieves 84.04% accuracy, outperforming the baseline model. The ROC shows tradeoffs between sensitivity and specificity, and AUC indicates the model can differentiate responsive from non-responsive emails 83.57% of the time.