#!/usr/bin/env python
# coding: utf-8

# # Installation

# In[ ]:


# run this cell to install pycaret in Google Colab
# !pip install pycaret


# In[ ]:


# If you are using Jupyter notebook, you can pip install pycaret using jupyter notebook or command line
# pip install pycaret


# In[3]:


from pycaret.utils import version
version()


# # 1. Importing Dataset

# In[5]:


from pycaret.datasets import get_data
data = get_data('juice')


# # 2. Setting up Environment

# In[10]:


from pycaret.classification import *
clf1 = setup(data, target = 'Purchase', session_id=786)


# # 3. Compare Models

# In[11]:


compare_models()


# # 4. Create Model

# In[12]:


lr = create_model('lr')


# In[13]:


dt = create_model('dt')


# # 5. Tune Model

# In[14]:


tuned_dt = tune_model('dt')


# In[19]:


print(tuned_dt)


# In[18]:


tuned_nb = tune_model('nb', optimize = 'AUC')


# In[20]:


print(tuned_nb)


# # 6. Ensemble Model

# In[21]:


bagged_dt = ensemble_model(dt)


# In[22]:


boosted_dt = ensemble_model(dt, method = 'Boosting')


# # 7. Blend Models

# In[23]:


lr = create_model('lr', verbose=False)
lda = create_model('lda', verbose=False)
gbc = create_model('gbc', verbose=False)


# In[25]:


blender = blend_models(estimator_list=[lr,lda,gbc], method = 'soft')


# In[26]:


blender.estimators_


# # 8. Analyze Model

# In[27]:


plot_model(blender)


# In[41]:


plot_model(blender, plot = 'confusion_matrix')


# In[42]:


plot_model(blender, plot = 'threshold')


# In[43]:


plot_model(blender, plot = 'pr')


# In[44]:


plot_model(tuned_dt, plot = 'vc')


# In[29]:


plot_model(dt, plot = 'boundary')


# In[31]:


plot_model(tuned_nb, plot = 'boundary')


# In[30]:


plot_model(blender, plot = 'boundary')


# In[33]:


evaluate_model(tuned_nb)


# # 9. Interpret Model

# In[38]:


xgboost = create_model('xgboost')


# In[39]:


interpret_model(xgboost)


# In[46]:


interpret_model(xgboost, plot = 'correlation')


# In[40]:


interpret_model(xgboost, plot = 'reason', observation=1)


# In[45]:


interpret_model(xgboost, plot = 'reason')


# # Learning Resources:
# 
# - PyCaret Classification Module : https://www.pycaret.org/classification
# - Binary Classification Tutorial (Level Beginner) : https://pycaret.org/clf101/
# - Binary Classification Tutorial (Level Intermediate) : https://pycaret.org/clf102/
# - Kaggle Titanic Predictions (Video Tutorial) : https://www.youtube.com/watch?v=nqMM6rngNCA