In gridsearchCV, when I fit like something as follows:
forest_reg = RandomForestRegressor()
grid_search = GridSearchCV(forest_reg, param_grid,cv=5,scoring = 'neg_mean_squared_error')
grid_search.fit(X_train,y_train)
and after that,
when I execute this,
GridSearch.best_estimator_.feature_importances_
it gives an array of values
so my question is what values does GridSearch.best_estimator_.feature_importances_ this line return??
In your case, GridSearch.best_estimator_.feature_importances_ returns a RandomForestRegressor object.
Therefore, according to RandomForestRegressor documentation:
feature_importances_ : array of shape = [n_features]
Return the feature importances (the higher, the more important the feature).
In other words, it returns the most important features according to your training set X_train. Each element of feature_importances_ corresponds to one feature of X_train (e.g: first element of feature_importances_ refers to the first feature/column of X_train).
The higher the value of an element in feature_importances_, the more important is the feature in X_train.
Related
I am trying to use GridSearchCV to optimize a pipeline that does feature selection in the beginning and classification using KNN at the end. I have fitted the model using my data set but when I see the best parameters found by GridSearchCV, it only gives the best parameters for SelectKBest. I have no idea why it doesn't show the best parameters for KNN.
Here is my code.
Addition of KNN and SelectKbest
classifier = KNeighborsClassifier()
parameters = {"classify__n_neighbors": list(range(5,15)),
"classify__p":[1,2]}
sel = SelectKBest(f_classif)
param={'kbest__k': [10, 20 ,30 ,40 ,50]}
GridsearchCV with pipeline and parameter grid
model = GridSearchCV(Pipeline([('kbest',sel),('classify', classifier)]),
param_grid=[param,parameters], cv=10)
fitting the model
model.fit(X_new, y)
the result
print(model.best_params_)
{'kbest__k': 40}
That's an incorrect way of merging dicts I believe. Try
param_grid={**param,**parameters}
or (Python 3.9+)
param_grid=param|parameters
When param_grid is a list, the disjoint union of the grids generated by each dictionary in the list is explored. So your search is over (1) the default k=10 selected features and every combination of classifier parameters, and separately (2) the default classifier parameters and each value of k. That the best parameters just show k=40 means that having more features, even with default classifier, performed best. You can check your cv_results_ to verify.
As dx2-66 answers, merging the dictionaries will generate the full grid you probably are after. You could also just define a single dictionary from the start.
Unless I am mistaken, a logistic regression model is simply a polynomial (typically where terms are to the first degree), where each attribute is a variable, and there is a weight associated with each variable.
I have trained a logistic regression model in Sci-Kit Learn, and would like to view the polynomial that represents the model, but I am unsure how to.
For the sake of elaborating with an example, suppose I have a dataset X with 4 attributes, and the corresponding binary labels y. Then, using the code below
clf = LogisticRegression(penalty='l2', tol=0.0001, solver='lbfgs', max_iter=1000)
model = clf.fit(X, y)
I would like to see the equation that represents the model. In other words, I want something like this, where the i'th alpha term is the weight to the i'th attribute of X, and beta is some constant.
I viewed the official documentation here, but did not see a way to do what I was hoping for. Are there any ideas on how I may get started doing what I want to do?
Like the comments already pointed out, you can get the weights (or coefficients in LogisticRegression) and the bias term (or intercept in LogisticRegression) via the coef_ and intercept_ attributes respectively. Here are the descriptions for both from the documentation:
coef_: ndarray of shape (1, n_features) or (n_classes, n_features)
Coefficient of the features in the decision function.
intercept_: ndarray of shape (1,) or (n_classes,).
Intercept (a.k.a. bias) added to the decision function.
Following your example, you could do something like this:
clf = LogisticRegression(penalty='l2', tol=0.0001, solver='lbfgs', max_iter=1000)
clf.fit(X, y)
print(clf.coef_)
print(clf.intercept_)
The values in coef_ correspond to the features, i.e. the ith value in coef_ corresponds to the ith feature in your training data.
I am using the LogisticRegression() method in scikit-learn on a highly unbalanced data set. I have even turned the class_weight feature to auto.
I know that in Logistic Regression it should be possible to know what is the threshold value for a particular pair of classes.
Is it possible to know what the threshold value is in each of the One-vs-All classes the LogisticRegression() method designs?
I did not find anything in the documentation page.
Does it by default apply the 0.5 value as threshold for all the classes regardless of the parameter values?
There is a little trick that I use, instead of using model.predict(test_data) use model.predict_proba(test_data). Then use a range of values for thresholds to analyze the effects on the prediction;
pred_proba_df = pd.DataFrame(model.predict_proba(x_test))
threshold_list = [0.05,0.1,0.15,0.2,0.25,0.3,0.35,0.4,0.45,0.5,0.55,0.6,0.65,.7,.75,.8,.85,.9,.95,.99]
for i in threshold_list:
print ('\n******** For i = {} ******'.format(i))
Y_test_pred = pred_proba_df.applymap(lambda x: 1 if x>i else 0)
test_accuracy = metrics.accuracy_score(Y_test.as_matrix().reshape(Y_test.as_matrix().size,1),
Y_test_pred.iloc[:,1].as_matrix().reshape(Y_test_pred.iloc[:,1].as_matrix().size,1))
print('Our testing accuracy is {}'.format(test_accuracy))
print(confusion_matrix(Y_test.as_matrix().reshape(Y_test.as_matrix().size,1),
Y_test_pred.iloc[:,1].as_matrix().reshape(Y_test_pred.iloc[:,1].as_matrix().size,1)))
Best!
Logistic regression chooses the class that has the biggest probability. In case of 2 classes, the threshold is 0.5: if P(Y=0) > 0.5 then obviously P(Y=0) > P(Y=1). The same stands for the multiclass setting: again, it chooses the class with the biggest probability (see e.g. Ng's lectures, the bottom lines).
Introducing special thresholds only affects in the proportion of false positives/false negatives (and thus in precision/recall tradeoff), but it is not the parameter of the LR model. See also the similar question.
Yes, Sci-Kit learn is using a threshold of P>=0.5 for binary classifications. I am going to build on some of the answers already posted with two options to check this:
One simple option is to extract the probabilities of each classification using the output from model.predict_proba(test_x) segment of the code below along with class predictions (output from model.predict(test_x) segment of code below). Then, append class predictions and their probabilities to your test dataframe as a check.
As another option, one can graphically view precision vs. recall at various thresholds using the following code.
### Predict test_y values and probabilities based on fitted logistic
regression model
pred_y=log.predict(test_x)
probs_y=log.predict_proba(test_x)
# probs_y is a 2-D array of probability of being labeled as 0 (first
column of
array) vs 1 (2nd column in array)
from sklearn.metrics import precision_recall_curve
precision, recall, thresholds = precision_recall_curve(test_y, probs_y[:,
1])
#retrieve probability of being 1(in second column of probs_y)
pr_auc = metrics.auc(recall, precision)
plt.title("Precision-Recall vs Threshold Chart")
plt.plot(thresholds, precision[: -1], "b--", label="Precision")
plt.plot(thresholds, recall[: -1], "r--", label="Recall")
plt.ylabel("Precision, Recall")
plt.xlabel("Threshold")
plt.legend(loc="lower left")
plt.ylim([0,1])
we can use a wrapper as follows:
model = LogisticRegression()
model.fit(X, y)
def custom_predict(X, threshold):
probs = model.predict_proba(X)
return (probs[:, 1] > threshold).astype(int)
new_preds = custom_predict(X=X, threshold=0.4)
I think I understand that until recently people used the attribute coef_ to extract the most informative features from linear models in python's machine learning library sklearn. Now users get pointed to SelectFromModel instead. SelectFromModel allows to reduce the features based on a threshold. So something like the following code reduces the features down to those features which have an importance > 0.5. My question now: Is there any way to determine whether a feature is positivly or negatively discriminating for a class?
I have my data in a pandas dataframe called data, first column a list of filenames of text files, second column the label.
count_vect = CountVectorizer(input="filename", analyzer="word")
X_train_counts = count_vect.fit_transform(data["filenames"])
print(X_train_counts.shape)
tf_transformer = TfidfTransformer(use_idf=True)
traindata = tf_transformer.fit_transform(X_train_counts)
print(traindata.shape) #report size of the training data
clf = LogisticRegression()
model = SelectFromModel(clf, threshold=0.5)
X_transform = model.fit_transform(traindata, data["labels"])
print("reduced features: ", X_transform.shape)
#get the names of all features
words = np.array(count_vect.get_feature_names())
#get the names of the important features using the boolean index from model
print(words[model.get_support()])
To my knowledge you need to stick back to the .coef_ method and see which coefficients are negative or positive. a negative coefficient obviously decreases the odds of that class to happen (so negative relationship), while a positive coefficient increases the odds the class to happen (so positive relationship).
However this method will not give you the significance, only the direction. You will need the SelectFromModel method to extract that.
I am trying to do features selection as a part of the a scikit-learn pipeline, on a multi-label scenario. My purpose is to select best K features, for some given k.
It might be simple, but I don't understand how to get the selected features indices in such a scenario.
on a regular scenario I could do something like that:
anova_filter = SelectKBest(f_classif, k=10)
anove_filter.fit_transform(data.X, data.Y)
anova_filter.get_support()
but on a multilabel scenario my labels dimensions are #samples X #unique_labels so fit and fit_transform yield the following exception:
ValueError: bad input shape
which makes sense, because it expects labels of dimension [#samples]
on the multilabel scenario, it makes sense to do something like that:
clf = Pipeline([('f_classif', SelectKBest(f_classif, k=10)),('svm', LinearSVC())])
multiclf = OneVsRestClassifier(clf, n_jobs=-1)
multiclf.fit(data.X, data.Y)
but then the object I'm getting is of type sklearn.multiclass.OneVsRestClassifier which doesn't have a get_support function. How do I get the trained SelectKBest model when it's used during a pipeline?
The way you set it up, there will be one SelectKBest per class. Is that what you intended?
You can get them via
multiclf.estimators_[i].named_steps['f_classif'].get_support()
If you want one feature selection for all the OvR models,
you can do
clf = Pipeline([('f_classif', SelectKBest(f_classif, k=10)),
('svm', OneVsRestClassifier(LinearSVC()))])
and get the single feature selection with
clf.named_steps['f_classif'].get_support()