def deserialize_random_forest_regressor(model_dict):
model = RandomForestRegressor(**model_dict['params'])
estimators = [
deserialize_decision_tree_regressor(decision_tree)
for decision_tree in model_dict['estimators_']
]
model.estimators_ = np.array(estimators)
model.n_features_ = model_dict['n_features_']
model.n_outputs_ = model_dict['n_outputs_']
model.max_depth = model_dict['max_depth']
model.min_samples_split = model_dict['min_samples_split']
model.min_samples_leaf = model_dict['min_samples_leaf']
model.min_weight_fraction_leaf = model_dict['min_weight_fraction_leaf']
model.max_features = model_dict['max_features']
model.max_leaf_nodes = model_dict['max_leaf_nodes']
model.min_impurity_decrease = model_dict['min_impurity_decrease']
model.min_impurity_split = model_dict['min_impurity_split']
if 'oob_score_' in model_dict:
model.oob_score_ = model_dict['oob_score_']
if 'oob_prediction_' in model_dict:
model.oob_prediction_ = np.array(model_dict['oob_prediction_'])
return model
def RFR(x_train,y_train,x_test,udf_trees=100,udf_max_features='auto', udf_min_samples=1, do_CV=False,names=None):
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error
from sklearn.cross_validation import cross_val_score
if do_CV:
### Randomly split up training set into 80/20 split. ###
### 80 for CV, 20 for "Test" score ###
from sklearn.cross_validation import train_test_split
x_train_cv, x_test_cv, y_train_cv, y_test_cv = train_test_split(x_train,y_train,test_size=0.20, random_state=42)
param_grid = {'max_features': [4,5,6],
'min_samples_leaf':[50,250,1000,2500]}
est=RandomForestRegressor(n_estimators=100,verbose=1, n_jobs=-1)
cv_scores=list()
test_scores=list()
params_list=list()
start = time()
for mfeatures in param_grid['max_features']:
for minSamples in param_grid['min_samples_leaf']:
print 'Trying parameter combination with 100 trees: (MaxFeatures=%i, minSamples=%i)' % (mfeatures,minSamples)
est.min_samples_leaf=minSamples
est.max_features=mfeatures
cv_score=cross_val_score(est,x_train_cv,y_train_cv,scoring='mean_squared_error',cv=5)
cv_scores.append(np.mean(cv_score))
### Create the labels for display purposes ###
params_list.append((mfeatures,minSamples))
### Perform 20% test set score ###
est.fit(x_train_cv,y_train_cv)
y_pred=est.predict(x_test_cv)
test_scores.append(mean_squared_error(y_test_cv,y_pred))
print 'Took %.2f seconds for parameter tuning.' %(time()-start)
print 'writing CV results to file...'
results = np.array([params_list,cv_scores,test_scores]).T ## should have 48 results...
print 'Parameter tuning results........'
print 'Parameters (max_features, min_samples_leaf), CV_Scores'
for i in range(len(results)):
print results[i]
else:
### Train the RFC Classifier with the optimal parameters found above ###
### RFR only takes 'MSE', need to change it to RMSEPE as per contest rules ###
print 'Fitting Random Forest with optimal user-defined parameters....'
est=RandomForestRegressor(n_estimators=udf_trees, max_features=udf_max_features,min_samples_leaf=udf_min_samples,n_jobs=-1,verbose=1)
est.fit(x_train,y_train)
#idx=np.where(x_test[:,1]==0)
#x_test=np.delete(x_test, 1, axis=1)
y_pred=est.predict(x_test)
y_pred=np.exp(y_pred)
#y_pred[idx] = 0
### Plot feature importances ###
#plot_feature_importance(est, names)
print 'Writing submission file....'
with open('RFC_Submission.csv','wb') as testfile:
w=csv.writer(testfile)
w.writerow(('Id','Sales'))
for i in range(len(y_pred)):
w.writerow(((i+1),y_pred[i]))
testfile.close()
print 'File written to disk...'
