From cf8be16806495f4bf2c3cf6b9b19612b486e94a2 Mon Sep 17 00:00:00 2001
From: moink <theresa.robinson@gmail.com>
Date: Tue, 7 May 2019 08:35:04 +0200
Subject: [PATCH] Fixed setup.py. Added to readme how to run scripts

---
 README.md       | 59 +++++++++++++++++++++++++++++++++++++++++++++++++
 run_epiclass.py |  2 +-
 setup.py        |  5 +++--
 3 files changed, 63 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index 095421b..ec0ea45 100644
--- a/README.md
+++ b/README.md
@@ -182,6 +182,65 @@ In order to demonstrate how one might deploy the model, I wrote a [web API](api.
 
 I also wrote a [few tests to ensure the models are working](/test_deployment.py) They only test the PCA-SVM models since these were the best models I developed. For each I test a single prediction, and then test that the confusion matrix has not changed. 
 
+## Working with epiclass
+
+### Installing
+
+Epiclass requires python 3. To get epiclass, make sure your python is python 3, and run 
+
+`git clone https://github.com/moink/epiclass/`
+
+`cd epiclass`
+
+`python setup.py`
+
+### Running
+
+There are three scripts to run, run_epiclass.py, api.py, and test_deployment.py
+
+#### run_epiclass.py
+
+This script is included to show how I created the plots in this readme, trained the models, and tuned the hyperparameters. Running it with all options may take a long time, depending on your hardware -- on the order of hours.
+
+To run it, provide one or more of the following actions as a parameter to the python script:
+
+* explore - make several plots of the data, including of the
+                          Principal Component Analysis (PCA) transformation of
+                          the features
+* pca_svm2 - train a binary classifier (seizure vs
+                          non-seizure) using a pipeline of PCA and a support 
+                          vector machine
+* pca_svm5 - train a multiclass classifier using a pipeline 
+                          of PCA and a support vector machine
+* rf - train a multiclass classifier using a random decision 
+                           forest
+* nn - train a multiclass classifier using an artificial 
+                           neural network
+                  All the training methods save the models to the model 
+                  directory and additionally save a confusion matrix to the 
+                  outputs directory.
+
+For example, to create the data exploration plots and train the neural network, run
+
+`python run_epiclass.py explore nn`
+
+#### test_deployment.py
+
+This runs four tests to ensure that the PCA-SVM models are giving expected results. To run it run:
+
+`python test_deployment.py`
+
+It will show some text in the console indicating whether the tests ran.
+
+#### api.py
+
+This starts a process serving an API with access to the binary PCA-SVM model.
+To run it run:
+
+`python api.py`
+
+Your python will give you a URL - visit it with your browser.
+
 ## Conclusion
 
 I think I could improve the models with further tuning (for example, a finer grid search) and by trying other models (e.g. different neural network architectures, logistic regression, fourier transforms). However I think the improvements would be marginal and that I have gotten pretty close to the limits of what the data can provide.
diff --git a/run_epiclass.py b/run_epiclass.py
index 24278a0..296969c 100644
--- a/run_epiclass.py
+++ b/run_epiclass.py
@@ -2,7 +2,7 @@
 
 # import modules used here -- sys is a very standard one
 import sys, argparse, logging
-from epiclass import run
+from epiclass.epiclass import run
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(
diff --git a/setup.py b/setup.py
index b9f6116..7540371 100644
--- a/setup.py
+++ b/setup.py
@@ -3,13 +3,14 @@
 setup(
     name='epiclass',
     version='1.0.0',
-    packages=['.'],
+    packages=['epiclass'],
     url='https://github.com/moink/epiclass',
     license='MIT',
     author='moink',
     author_email='',
     description='Visualization and prediction of epileptic seizure data set',
     install_requires=['keras', 'pandas', 'joblib', 'matplotlib', 'seaborn',
-                      'scikit-learn', 'flask', 'flask_restful', 'tensorflow'],
+                      'scikit-learn', 'flask>=1.0.2', 'flask_restful',
+                      'tensorflow'],
     scripts=['run_epiclass', 'api', 'test_deployment']
 )
\ No newline at end of file