I have some data (MFCC features for speaker recognition), from two different speakers. 60 vectors of 13 features for each person (in total 120). Each of them has their label (0 and 1). I need to show the results on confusion matrix. But GaussianMixture model from sklearn is unstable. For each program run i receive different scores (sometimes accuracy is 0.4, sometimes 0.7 ...). I don't know what I am doing wrong, because analogically i created SVM and k-NN models and they are working fine (stable accuracy around 0.9). Do you have any idea what am I doing wrong?
gmmclf = GaussianMixture(n_components=2, covariance_type='diag')
gmmclf.fit(X_train, y_train) #X_train are mfcc vectors, y_train are labels
ygmm_pred_class = gmmclf.predict(X_test)
print(accuracy_score(y_test, ygmm_pred_class))
print(confusion_matrix(y_test, ygmm_pred_class))
Short answer: you should simply not use a GMM for classification.
Long answer...
From the answer to a relevant thread, Multiclass classification using Gaussian Mixture Models with scikit learn (emphasis in the original):
Gaussian Mixture is not a classifier. It is a density estimation
method, and expecting that its components will magically align with
your classes is not a good idea. [...] GMM simply tries to fit mixture of Gaussians
into your data, but there is nothing forcing it to place them
according to the labeling (which is not even provided in the fit
call). From time to time this will work - but only for trivial
problems, where classes are so well separated that even Naive Bayes
would work, in general however it is simply invalid tool for the
problem.
And a comment by the respondent himself (again, emphasis in the original):
As stated in the answer - GMM is not a classifier, so asking if you
are using "GMM classifier" correctly is impossible to answer. Using
GMM as a classifier is incorrect by definition, there is no "valid"
way of using it in such a problem as it is not what this model is
designed to do. What you could do is to build a proper generative
model per class. In other words construct your own classifier where
you fit one GMM per label and then use assigned probability to do
actual classification. Then it is a proper classifier. See
github.com/scikit-learn/scikit-learn/pull/2468
(For what it may worth, you may want to notice that the respondent is a research scientist in DeepMind, and the very first person to be awarded the machine-learning gold badge here at SO)
To elaborate further (and that's why I didn't simply flag the question as a duplicate):
It is true that in the scikit-learn documentation there is a post titled GMM classification:
Demonstration of Gaussian mixture models for classification.
which I guess did not exist back in 2017, when the above response was written. But, digging into the provided code, you will realize that the GMM models are actually used there in the way proposed by lejlot above; there is no statement in the form of classifier.fit(X_train, y_train) - all usage is in the form of classifier.fit(X_train), i.e. without using the actual labels.
This is exactly what we would expect from a clustering-like algorithm (which is indeed what GMM is), and not from a classifier. It is true again that scikit-learn offers an option for providing also the labels in the GMM fit method:
fit (self, X, y=None)
which you have actually used here (and again, probably did not exist back in 2017, as the above response implies), but, given what we know about GMMs and their usage, it is not exactly clear what this parameter is there for (and, permit me to say, scikit-learn has its share on practices that may look sensible from a purely programming perspective, but which made very little sense from a modeling perspective).
A final word: although fixing the random seed (as suggested in a comment) may appear to "work", trusting a "classifier" that gives a range of accuracies between 0.4 and 0.7 depending on the random seed is arguably not a good idea...
In sklearn, the labels of clusters in gmm do not mean anything. So, each time you run a gmm, the labels may vary. It might be one reason the results are not robust.
Related
I have a highly unbalanced dataset of 3 classes. To address this, I applied the sample_weight array in the XGBClassifier, but I'm not noticing any changes in the modelling results? All of the metrics in the classification report (confusion matrix) are the same. Is there an issue with the implementation?
The class ratios:
military: 1171
government: 34852
other: 20869
Example:
pipeline = Pipeline([
('bow', CountVectorizer(analyzer=process_text)), # convert strings to integer counts
('tfidf', TfidfTransformer()), # convert integer counts to weighted TF-IDF scores
('classifier', XGBClassifier(sample_weight=compute_sample_weight(class_weight='balanced', y=y_train))) # train on TF-IDF vectors w/ Naive Bayes classifier
])
Sample of Dataset:
data = pd.DataFrame({'entity_name': ['UNICEF', 'US Military', 'Ryan Miller'],
'class': ['government', 'military', 'other']})
Classification Report
First, most important: use a multiclass eval_metric. eval_metric=merror or mlogloss, then post us the results. You showed us ['precision','recall','f1-score','support'], but that's suboptimal, or outright broken unless you computed them in a multi-class-aware, imbalanced-aware way.
Second, you need weights. Your class ratio is military: government: other 1:30:18, or as percentages 2:61:37%.
You can manually set per-class weights with xgb.DMatrix..., weights)
Look inside your pipeline (use print or verbose settings, dump values), don't just blindly rely on boilerplate like sklearn.utils.class_weight.compute_sample_weight('balanced', ...) to give you optimal weights.
Experiment with manually setting per-class weights, starting with 1 : 1/30 : 1/18 and try more extreme values. Reciprocals so the rarer class gets higher weight.
Also try setting min_child_weight much higher, so it requires a few exemplars (of the minority classes). Start with min_child_weight >= 2(* weight of rarest class) and try going higher. Beware of overfitting to the very rare minority class (this is why people use StratifiedKFold crossvalidation, for some protection, but your code isn't using CV).
We can't see your other parameters for xgboost classifier (how many estimators? early stopping on or off? what was learning_rate/eta? etc etc.). Seems like you used the defaults - they'll be terrible. Or else you're not showing your code. Distrust xgboost's defaults, esp. for multiclass, don't expect xgboost to give good out-of-the-box results. Read the doc and experiment with values.
Do all that experimentation, post your results, check before concluding "it doesn't work". Don't expect optimal results from out-of-the-box. Distrust or double-check the sklearn util functions, try manual alternatives. (Often, just because sklearn has a function to do something, doesn't mean it's good or best or suitable for all use-cases, like imbalanced multiclass)
I am dealing with a multi-class problem (4 classes) and I am trying to solve it with scikit-learn in Python.
I saw that I have three options:
I simply instantiate a classifier, then I fit with train and evaluate with test;
classifier = sv.LinearSVC(random_state=123)
classifier.fit(Xtrain, ytrain)
classifier.score(Xtest, ytest)
I "encapsulate" the instantiated classifier in a OneVsRest object, generating a new classifier that I use for train and test;
classifier = OneVsRestClassifier(svm.LinearSVC(random_state=123))
classifier.fit(Xtrain, ytrain)
classifier.score(Xtest, ytest)
I "encapsulate" the instantiated classifier in a OneVsOne object, generating a new classifier that I use for train and test.
classifier = OneVsOneClassifier(svm.LinearSVC(random_state=123))
classifier.fit(Xtrain, ytrain)
classifier.score(Xtest, ytest)
I understand the difference between OneVsRest and OneVsOne, but I cannot understand what I am doing in the first scenario where I do not explicitly pick up any of these two options. What does scikit-learn do in that case? Does it implicitly use OneVsRest?
Any clarification on the matter would be highly appreciated.
Best,
MR
Edit:
Just to make things clear, I am not specifically interested in the case of SVMs. For example, what about RandomForest?
Updated answer: As clarified in the comments and edits, the question is more about the general setting of sklearn, and less about the specific case of LinearSVC which is explained below.
The main difference here is that some of the classifiers you can use have "built-in multiclass classification support", i.e. it is possible for that algorithm to discern between more than two classes by default. One example for this would for example be a Random Forest, or a Multi-Layer Perceptron (MLP) with multiple output nodes.
In these cases, having a OneVs object is not required at all, since you are already solving your task. In fact, using such a strategie might even decreaes your performance, since you are "hiding" potential correlations from the algorithm, by letting it only decide between single binary instances.
On the other hand, algorithms like SVC or LinearSVC only support binary classification. So, to extend these classes of (well-performing) algorithms, we instead have to rely on the reduction to a binary classification task, from our initial multiclass classification task.
As far as I am aware of, the most complete overview can be found here:
If you scroll down a little bit, you can see which one of the algorithms is inherently multiclass, or uses either one of the strategies by default.
Note that all of the listed algorithms under OVO actually now employ a OVR strategy by default! This seems to be slightly outdated information in that regard.
Initial answer:
This is a question that can easily be answered by looking at the relevant scikit-learn documentation.
Generally, the expectation on Stackoverflow is that you have at least done some form of research on your own, so please consider looking into existing documentation first.
multi_class : string, ‘ovr’ or ‘crammer_singer’ (default=’ovr’)
Determines the multi-class strategy if y contains more than two classes. "ovr" trains n_classes one-vs-rest classifiers, while
"crammer_singer" optimizes a joint objective over all classes. While
crammer_singer is interesting from a theoretical perspective as it is
consistent, it is seldom used in practice as it rarely leads to better
accuracy and is more expensive to compute. If "crammer_singer" is
chosen, the options loss, penalty and dual will be ignored.
So, clearly, it uses one-vs-rest.
The same holds by the way for the "regular" SVC.
I am working with a dataset of about 400.000 x 250.
I have a problem with the model yielding a very good R^2 score when testing it on the training set, but extremely poorly when used on the test set. Initially, this sounds like overfitting. But the data is split into training/test set at random and the data set i pretty big, so I feel like there has to be something else.
Any suggestions?
Splitting dataset into training set and test set
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(df.drop(['SalePrice'],
axis=1), df.SalePrice, test_size = 0.3)
Sklearn's Linear Regression estimator
from sklearn import linear_model
linReg = linear_model.LinearRegression() # Create linear regression object
linReg.fit(X_train, y_train) # Train the model using the training sets
# Predict from training set
y_train_linreg = linReg.predict(X_train)
# Predict from test set
y_pred_linreg = linReg.predict(X_test)
Metric calculation
from sklearn import metrics
metrics.r2_score(y_train, y_train_linreg)
metrics.r2_score(y_test, y_pred_linreg)
R^2 score when testing on training set: 0,64
R^2 score when testing on testing set: -10^23 (approximatly)
While I agree with Mihai that your problem definitely looks like overfitting, I don't necessarily agree on his answer that neural network would solve your problem; at least, not out of the box. By themselves, neural networks overfit more, not less, than linear models. You need somehow to take care of your data, hardly any model can do that for you. A few options that you might consider (apologies, I cannot be more precise without looking at the dataset):
Easiest thing, use regularization. 400k rows is a lot, but with 250 dimensions you can overfit almost whatever you like. So try replacing LinearRegression by Ridge or Lasso (or Elastic Net or whatever). See http://scikit-learn.org/stable/modules/linear_model.html (Lasso has the advantage of discarding features for you, see next point)
Especially if you want to go outside of linear models (and you probably should), it's advisable to first reduce the dimension of the problem, as I said 250 is a lot. Try using some of the Feature selection techniques here: http://scikit-learn.org/stable/modules/feature_selection.html
Probably most importantly than anything else, you should consider adapting your input data. The very first thing I'd try is, assuming you are really trying to predict a price as your code implies, to replace it by its logarithm, or log(1+x). Otherwise linear regression will try very very hard to fit that single object that was sold for 1 Million $ ignoring everything below $1k. Just as important, check if you have any non-numeric (categorical) columns and keep them only if you need them, in case reducing them to macro-categories: a categorical column with 1000 possible values will increase your problem dimension by 1000, making it an assured overfit. A single column with a unique categorical data for each input (e.g. buyer name) will lead you straight to perfect overfitting.
After all this (cleaning data, reducing dimension via either one of the methods above or just Lasso regression until you get to certainly less than dim 100, possibly less than 20 - and remember that this includes any categorical data!), you should consider non-linear methods to further improve your results - but that's useless until your linear model provides you at least some mildly positive R^2 value on test data. sklearn provides a lot of them: http://scikit-learn.org/stable/modules/kernel_ridge.html is the easiest to use out-of-the-box (also does regularization), but it might be too slow to use in your case (you should first try this, and any of the following, on a subset of your data, say 1000 rows once you've selected only 10 or 20 features and see how slow that is). http://scikit-learn.org/stable/modules/svm.html#regression have many different flavours, but I think all but the linear one would be too slow. Sticking to linear things, http://scikit-learn.org/stable/modules/sgd.html#regression is probably the fastest, and would be how I'd train a linear model on this many samples. Going truly out of linear, the easiest techniques would probably include some kind of trees, either directly http://scikit-learn.org/stable/modules/tree.html#regression (but that's an almost-certain overfit) or, better, using some ensemble technique (random forests http://scikit-learn.org/stable/modules/ensemble.html#forests-of-randomized-trees are the typical go-to algorithm, gradient boosting http://scikit-learn.org/stable/modules/ensemble.html#gradient-tree-boosting sometimes works better). Finally, state-of-the-art results are indeed generally obtained via neural networks, see e.g. http://scikit-learn.org/stable/modules/neural_networks_supervised.html but for these methods sklearn is generally not the right answer and you should take a look at dedicated environments (TensorFlow, Caffe, PyTorch, etc.)... however if you're not familiar with those it is certainly not worth the trouble!
Use case:
I have a small dataset with about 3-10 samples in each class. I am using sklearn SVC to classify those with rbf kernel.
I need the confidence of the prediction along with the predicted class. I used predict_proba method of SVC.
I was getting weird results with that. I searched a bit and found out that it makes sense only for larger datasets.
Found this question on stack Scikit-learn predict_proba gives wrong answers.
The author of the question verified this by multiplying the dataset, thereby duplicating the dataset.
My questions:
1) If I multiply my dataset by lets say 100, having each sample 100 times, it increases the "correctness" of "predict_proba". What sideeffects will it have? Overfitting?
2) Is there any other way I can calculate the confidence of the classifier? Like distance from the hyperplanes?
3) For this small sample size, is SVM a recommended algorithm or should I choose something else?
First of all: Your data set seems very small for any practical purposes. That being said, let's see what we can do.
SVM's are mainly popular in high dimensional settings. It is currently unclear whether that applies to your project. They build planes on a handful of (or even single) supporting instances, and are often outperformed in situation with large trainingsets by Neural Nets. A priori they might not be your worse choice.
Oversampling your data will do little for an approach using SVM. SVM is based on the notion of support vectors, which are basically the outliers of a class that define what is in the class and what is not. Oversampling will not construct new support vector (I am assuming you are already using the train set as test set).
Plain oversampling in this scenario will also not give you any new information on confidence, other than artififacts constructed by unbalanced oversampling, since the instances will be exact copies and no distibution changes will occur. You might be able to find some information by using SMOTE (Synthetic Minority Oversampling Technique). You will basically generate synthetic instances based of the ones you have. In theory this will provide you with new instances, that won't be exact copies of the ones you have, and might thusly fall a little out of the normal classification. Note: By definition all these examples will lie in between the original examples in your sample space. This will not mean that they will lie in between your projected SVM-space, possibly learning effects that aren't really true.
Lastly, you can estimate confidence with the distance to the hyperplane. Please see: https://stats.stackexchange.com/questions/55072/svm-confidence-according-to-distance-from-hyperline
I'm trying to run a classifier in a set of about 1000 objects, each with 6 floating point variables. I've used scikit-learn's cross validation features to generate an array of the predicted values for several different models. I've then used sklearn.metrics to compute the accuracy of my classifiers, and the confusion table. Most classifiers have around 20-30% accuracy. Below is the confusion table for the SVC classifier (25.4% accuracy).
Since I'm new to machine learning, I'm not sure how to interpret that result, and whether there are other good metrics to evaluate the problem. Intuitively speaking, even with 25% accuracy, and given that the classifier got 25% of the predictions right, I believe it is at least somewhat effective, right? How can I express that with statistical arguments?
If this table is a confusion table, I think that your classifier predicts in majority of the time the class E. I think that your class E is overrepresented in your dataset, accuracy is not a good metric if your classes have not the same number of instances,
Example, If you have 3 classes, A,B,C and in the test dataset the class A is over represented (90%) if your classifier predicts all time class A, you will have 90% of accuracy,
A good metric is to use log loss, logistic regression is a good algorithm that optimize this metric
see https://stats.stackexchange.com/questions/113301/multi-class-logarithmic-loss-function-per-class
An other solution, is to do oversampling of your small classes
First of all, I find it very difficult to look at confusion tables. Plotting it as an image would give a lot better intuitive understanding about what is going on.
It is advisory to have single number metric to optimize since it is easier and faster. When you find that your system doesn't perform as you expect it to, revise your selection of metric.
Accuracy is usually a good metric to use if you have same amount of examples in every class. Otherwise (which seems to be the case here) I'd advise to use F1 score which takes into account both precision and recall of your estimator.
EDIT: However it is up to you to decide if the ~25% accuracy, or whatever metric is "good enough". If you are classifying if robot should shoot a person you should probably revise your algorithm but if you are deciding if it is a pseudo-random or random data, 25% percent accuracy could be more than enough to prove the point.