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

# In[ ]:





# In[36]:


import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error


# In[58]:


# metric de scoring
def log_rmse(yhat, ytrue):
    return np.sqrt(mean_squared_error ( np.log(yhat), np.log(ytrue) ))


# In[59]:


df = pd.read_csv('ames/train.csv')
df.head()


# In[60]:


# selection arbitraire des colonnes
# check for NaN
cols = ['MoSold', 'YrSold','1stFlrSF','2ndFlrSF', 'BedroomAbvGr']
for col in cols:
    print("{}: {}".format(col, df[df[col].isnull()].shape ))

# pas de null values dans ces colonnes
X = df[cols]
y = df['SalePrice']


# In[61]:


# split train test

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X,y,
    test_size=0.3,
    random_state=2)




# In[62]:


# train regressor


clf = RandomForestRegressor(max_depth = 5,n_estimators=300 )

clf.fit(X_train, y_train)
print(clf)
yhat_test = clf.predict(X_test)
yhat_train = clf.predict(X_train)

print("\n -- RMSE test {:.4f}".format(log_rmse(y_test, yhat_test)  ))
print("\n -- RMSE train {:.4f}".format(log_rmse(y_train, yhat_train)))


# # Appliquer le model sur le test

# In[63]:


# load validation dataset
vdf = pd.read_csv('ames/test.csv')
# a ce niveau appliquer les memes tranformations sur X_valid
X_valid = vdf[cols]

y_valid = clf.predict(X_valid)


# In[64]:


# construire la dataframe de resultats
results = pd.DataFrame(columns = ['Id', 'SalePrice'])

# Kaggle veut que la colonne index commence par 1461
results['Id'] = X_valid.index + 1461

results['SalePrice'] = y_valid

# ecrire le resultats dans le fichier csv
results.to_csv("submission_01.csv", index = False)


# In[ ]:





# In[ ]: