Future work:
By checking the result I found the dataset I have is not well balance, most of case can get closed within a day, and random forest model tends to classify the test data into the biggest group.  To better handle this unbalanced data I could assign different weights to different groups by giving smaller groups higher weights and bigger groups lower weights. Also I can over sampling some smaller group data record to increase their group size.
Links
Bibliography:
Random Forest in Python
Understanding Random Forests Classifiers in Python
Random Forest Algorithm with Python and Scikit-Learn
Classification and Regression by RandomForest
Classification of emergency situations
Machine Learning with Python: Random Forests in Python