Missouri State University: CSC 535/635 - Data Mining

Professor: Dr. Jamil Saquer

Final Project: Patient Length of Stay Classifier

Contributors: David Gray, Braden Bagby, & Binh Le

Data Source: https://www.kaggle.com/nehaprabhavalkar/av-healthcare-analytics-ii

Instructions

1. run preprocess.py

   • Running this will preprocess and save the data.
   • After running there will be files named X_test.csv, X_train.csv, y_test.csv, and y_train.csv in the 'processed_data' folder.

2. run train_mlp.py or train_knn.py

   • Running this will train and save the corresponding algorithm.
   • After running there will be a file named mlp.pkl or knn.pkl in the 'models' folder.

3. run test_mlp.py or test_knn.py

   • Running this will produce a confusion matrix and classification report with the algorithms accuracy, precision, recall, and f1-score.

Project Folders & Files Overview

• /code: contains the source code
  • tools.py: full of tools for training, testing, and visualizing
  • preprocess.py: preprocesses data
  • train_knn.py: trains the knn algorithm
  • train_mlp.py: trains the mlp algorithm
  • test_knn.py: tests the knn algorithm
  • test_mlp.py: tests the mlp algorithm
  • tsne.py: applies tsne to the data and saves the visual
• /premade_examples: contains images, trained algorithms, and preprocessed data
  • knn.pkl: pretrained knn algorithm
  • mlp.pkl pretrained mlp algorithm
  • tsne.PNG: example results from tsne.py
  • knn_cm.png: example results from test_knn.py
  • mlp_cm.png: example results from test_mlp.py
  • knn_scores.png: example results from test_knn.py
  • mlp_scores.png: example results from test_mlp.py
• data.csv: original, unmodified healthcare data set for training and testing

Dependencies

1. pandas