diff --git a/Machine_Learning_for_Asset_Managers/ch6_feature_importance_analysis.py b/Machine_Learning_for_Asset_Managers/ch6_feature_importance_analysis.py
index eea5d56..ae5c78d 100644
--- a/Machine_Learning_for_Asset_Managers/ch6_feature_importance_analysis.py
+++ b/Machine_Learning_for_Asset_Managers/ch6_feature_importance_analysis.py
@@ -10,7 +10,7 @@
 import statsmodels.api as sm1
 import matplotlib.pylab as plt
 
-from ch4_optimal_clustering import clusterKMeansBase 
+from .ch4_optimal_clustering import clusterKMeansBase 
 
 #Code snippet 6.1 generating a set of informative, redundant, and noisy explanatory variables
 # returns matrix X of training samples, and vector y of class labels for the training samples
@@ -30,7 +30,7 @@ def getTestData(n_features=100, n_informative=25, n_redundant=25, n_samples=1000
 #code snippet 6.2 implementation of an ensembke MDI method
 def featImpMDI(fit, featNames):
     #feat importance based on IS mean impurity reduction
-    df0 = {i:tree.feature_importances_ for i, tree in enumerate(fit.enumerators_)}
+    df0 = {i:tree.feature_importances_ for i, tree in enumerate(fit.estimators_)}
     df0 = pd.DataFrame.from_dict(df0, orient='index')
     df0.columns = featNames
     df0 = df0.replace(0, np.nan) #because max_features=1