I'm working with an extremelly unbalanced and heterogeneous multiclass {K = 16} database for research, with a small N ~= 250. For some labels the database has a sufficient amount of examples for supervised machine learning, but for others I have almost none. I'm also not in a position to expand my database for a number of reasons.
As a first approach I divided my database into training (80%) and test (20%) sets in a stratified way. On top of that, I applied several classification algorithms that provide some results. I applied this procedure over 500 stratified train/test sets (as each stratified sampling takes individuals randomly within each stratum), hoping to select an algorithm (model) that performed acceptably.
Because of my database, depending on the specific examples that are part of the train set, the performance on the test set varies greatly. I'm dealing with runs that have as high (for my application) as 82% accuracy and runs that have as low as 40%. The median over all runs is around 67% accuracy.
When facing this situation, I'm unsure on what is the standard procedure (if there is any) when selecting the best performing model. My rationale is that the 90% model may generalize better because the specific examples selected in the training set are be richer so that the test set is better classified. However, I'm fully aware of the possibility of the test set being composed of "simpler" cases that are easier to classify or the train set comprising all hard-to-classify cases.
Is there any standard procedure to select the best performing model considering that the distribution of examples in my train/test sets cause the results to vary greatly? Am I making a conceptual mistake somewhere? Do practitioners usually select the best performing model without any further exploration?
I don't like the idea of using the mean/median accuracy, as obviously some models generalize better than others, but I'm by no means an expert in the field.
Confusion matrix of the predicted label on the test set of one of the best cases:
Confusion matrix of the predicted label on the test set of one of the worst cases:
They both use the same algorithm and parameters.
Good Accuracy =/= Good Model
I want to firstly point out that a good accuracy on your test set need not equal a good model in general! This has (in your case) mainly to do with your extremely skewed distribution of samples.
Especially when doing a stratified split, and having one class dominatingly represented, you will likely get good results by simply predicting this one class over and over again.
A good way to see if this is happening is to look at a confusion matrix (better picture here) of your predictions.
If there is one class that seems to confuse other classes as well, that is an indicator for a bad model. I would argue that in your case it would be generally very hard to find a good model unless you do actively try to balance your classes more during training.
Use the power of Ensembles
Another idea is indeed to use ensembling over multiple models (in your case resulting from different splits), since it is assumed to generalize better.
Even if you might sacrifice a lot of accuracy on paper, I would bet that a confusion matrix of an ensemble is likely to look much better than the one of a single "high accuracy" model. Especially if you disregard the models that perform extremely poor (make sure that, again, the "poor" performance comes from an actual bad performance, and not just an unlucky split), I can see a very good generalization.
Try k-fold Cross-Validation
Another common technique is k-fold cross-validation. Instead of performing your evaluation on a single 80/20 split, you essentially divide your data in k equally large sets, and then always train on k-1 sets, while evaluating on the other set. You then not only get a feeling whether your split was reasonable (you usually get all the results for different splits in k-fold CV implementations, like the one from sklearn), but you also get an overall score that tells you the average of all folds.
Note that 5-fold CV would equal a split into 5 20% sets, so essentially what you are doing now, plus the "shuffling part".
CV is also a good way to deal with little training data, in settings where you have imbalanced classes, or where you generally want to make sure your model actually performs well.
Related
For a class project, I designed a neural network to approximate sin(x), but ended up with a NN that just memorized my function over the data points I gave it. My NN took in x-values with a batch size of 200. Each x-value was multiplied by 200 different weights, mapping to 200 different neurons in my first layer. My first hidden layer contained 200 neurons, each one a linear combination of the x-values in the batch. My second hidden layer also contained 200 neurons, and my loss function was computed between the 200 neurons in my second layer and the 200 values of sin(x) that the input mapped to.
The problem is, my NN perfectly "approximated" sin(x) with 0 loss, but I know it wouldn't generalize to other data points.
What did I do wrong in designing this neural network, and how can I avoid memorization and instead design my NN's to "learn" about the patterns in my data?
It is same with any machine learning algorithm. You have a dataset based on which you try to learn "the" function f(x), which actually generated the data. In real life datasets, it is impossible to get the original function from the data, and therefore we approximate it using something g(x).
The main goal of any machine learning algorithm is to predict unseen data as best as possible using the function g(x).
Given a dataset D you can always train a model, which will perfectly classify all the datapoints (you can use a hashmap to get 0 error on the train set), but which is overfitting or memorization.
To avoid such things, you yourself have to make sure that the model does not memorise and learns the function. There are a few things which can be done. I am trying to write them down in an informal way (with links).
Train, Validation, Test
If you have large enough dataset, use Train, Validation, Test splits. Split the dataset in three parts. Typically 60%, 20% and 20% for Training, Validation and Test, respectively. (These numbers can vary based on need, also in case of imbalanced data, check how to get stratified partitions which preserve the class ratios in every split). Next, forget about the Test partition, keep it somewhere safe, don't touch it. Your model, will be trained using the Training partition. Once you have trained the model, evaluate the performance of the model using the Validation set. Then select another set of hyper-parameter configuration for your model (eg. number of hidden layer, learaning algorithm, other parameters etc.) and then train the model again, and evaluate based on Validation set. Keep on doing this for several such models. Then select the model, which got you the best validation score.
The role of validation set here is to check what the model has learned. If the model has overfit, then the validation scores will be very bad, and therefore in the above process you will discard those overfit models. But keep in mind, although you did not use the Validation set to train the model, directly, but the Validation set was used indirectly to select the model.
Once you have selected a final model based on Validation set. Now take out your Test set, as if you just got new dataset from real life, which no one has ever seen. The prediction of the model on this Test set will be an indication how well your model has "learned" as it is now trying to predict datapoints which it has never seen (directly or indirectly).
It is key to not go back and tune your model based on the Test score. This is because once you do this, the Test set will start contributing to your mode.
Crossvalidation and bootstrap sampling
On the other hand, if your dataset is small. You can use bootstrap sampling, or k-fold cross-validation. These ideas are similar. For example, for k-fold cross-validation, if k=5, then you split the dataset in 5 parts (also be carefull about stratified sampling). Let's name the parts a,b,c,d,e. Use the partitions [a,b,c,d] to train and get the prediction scores on [e] only. Next, use the partitions [a,b,c,e] and use the prediction scores on [d] only, and continue 5 times, where each time, you keep one partition alone and train the model with the other 4. After this, take an average of these scores. This is indicative of that your model might perform if it sees new data. It is also a good practice to do this multiple times and perform an average. For example, for smaller datasets, perform a 10 time 10-folds cross-validation, which will give a pretty stable score (depending on the dataset) which will be indicative of the prediction performance.
Bootstrap sampling is similar, but you need to sample the same number of datapoints (depends) with replacement from the dataset and use this sample to train. This set will have some datapoints repeated (as it was a sample with replacement). Then use the missing datapoins from the training dataset to evaluate the model. Perform this multiple times and average the performance.
Others
Other ways are to incorporate regularisation techniques in the classifier cost function itself. For example in Support Vector Machines, the cost function enforces conditions such that the decision boundary maintains a "margin" or a gap between two class regions. In neural networks one can also do similar things (although it is not same as in SVM).
In neural network you can use early stopping to stop the training. What this does, is train on the Train dataset, but at each epoch, it evaluates the performance on the Validation dataset. If the model starts to overfit from a specific epoch, then the error for Training dataset will keep on decreasing, but the error of the Validation dataset will start increasing, indicating that your model is overfitting. Based on this one can stop training.
A large dataset from real world tends not to overfit too much (citation needed). Also, if you have too many parameters in your model (to many hidden units and layers), and if the model is unnecessarily complex, it will tend to overfit. A model with lesser pameter will never overfit (though can underfit, if parameters are too low).
In the case of you sin function task, the neural net has to overfit, as it is ... the sin function. These tests can really help debug and experiment with your code.
Another important note, if you try to do a Train, Validation, Test, or k-fold crossvalidation on the data generated by the sin function dataset, then splitting it in the "usual" way will not work as in this case we are dealing with a time-series, and for those cases, one can use techniques mentioned here
First of all, I think it's a great project to approximate sin(x). It would be great if you could share the snippet or some additional details so that we could pin point the exact problem.
However, I think that the problem is that you are overfitting the data hence you are not able to generalize well to other data points.
Few tricks that might work,
Get more training points
Go for regularization
Add a test set so that you know whether you are overfitting or not.
Keep in mind that 0 loss or 100% accuracy is mostly not good on training set.
I have a 6-dimensional training dataset where there is a perfect numeric attribute which separates all the training examples this way: if TIME<200 then the example belongs to class1, if TIME>=200 then example belongs to class2. J48 creates a tree with only 1 level and this attribute as the only node.
However, the test dataset does not follow this hypothesis and all the examples are missclassified. I'm having trouble figuring out whether this case is considered overfitting or not. I would say it is not as the dataset is that simple, but as far as I understood the definition of overfit, it implies a high fitting to the training data, and this I what I have. Any help?
However, the test dataset does not follow this hypothesis and all the examples are missclassified. I'm having trouble figuring out whether this case is considered overfitting or not. I would say it is not as the dataset is that simple, but as far as I understood the definition of overfit, it implies a high fitting to the training data, and this I what I have. Any help?
Usually great training score and bad testing means overfitting. But this assumes IID of the data, and you are clearly violating this assumption - your training data is completely different from the testing one (there is a clear rule for the training data which has no meaning for testing one). In other words - your train/test split is incorrect, or your whole problem does not follow basic assumptions of where to use statistical ml. Of course we often fit models without valid assumptions about the data, in your case - the most natural approach is to drop a feature which violates the assumption the most - the one used to construct the node. This kind of "expert decisions" should be done prior to building any classifier, you have to think about "what is different in test scenario as compared to training one" and remove things that show this difference - otherwise you have heavy skew in your data collection, thus statistical methods will fail.
Yes, it is an overfit. The first rule in creating a training set is to make it look as much like any other set as possible. Your training set is clearly different than any other. It has the answer embedded within it while your test set doesn't. Any learning algorithm will likely find the correlation to the answer and use it and, just like the J48 algorithm, will regard the other variables as noise. The software equivalent of Clever Hans.
You can overcome this by either removing the variable or by training on a set drawn randomly from the entire available set. However, since you know that there is a subset with an embedded major hint, you should remove the hint.
You're lucky. At times these hints can be quite subtle which you won't discover until you start applying the model to future data.
(noob in ML, be patient)
I want to test the performance of my scikit-learn SVMLinear classifier. My train-set has a different class distribution than the actual population, but my test-set is a representative, and distributes like the actual population.
I noticed that there's a class-weight parameter, and I want to try giving my classifier the actual population distribution, and see if it helps it perform better.
However - as my train-set distribution is different, so will be my validation set, right? So should I expect an improvement on the validation, or must I use my test-set to see the improvement? And if so - isn't it against the rules to calibrate using the test-set which will lead to burning the test-set or overfitting?
I've thought about bootstrap re-sampling of my train-set: making it distribute the same as the general population, and only then training and validating my model. Is this a good solution?
Thanks!
It seems that you have some good ideas which are mostly worth trying. The answers mostly depend on the application and the size of your train/test set.
It is against the rules to calibrate based on test set and again use the whole test set for evaluation. However, if your test set is large enough, you can always divide your test set to two sets: validation set, and actual test set. Then, your final evaluation will be based on a smaller test set, which might be still acceptable depending on the application.
For your training set that you believe it has different class distribution than the actual population, there might be several things worth trying. Usually the most acceptable approach is to use a classifier that can handle these differences (usually with fewer parameters to avoid over-fitting). There is a whole topic of classification and regression on skewed datasets that you can look through. Other than the classifier, provided that you did not derive the actual population from your test set, the methods below might help too:
1- One of them can be (as you said) bootstrap re-sampling in case that your training set is large enough for that.
2- Another approach can be generating more training samples by adding some noise to the current samples of the training set. For example if you are classifying images of birds, you can randomly make images darker or brighter, or randomly move them a few pixels to the sides or up and down (select values randomly in a small enough range). This way, you can add to the training set in a way to get the desired distribution.
I am doing a logistic regression to predict the outcome of a binary variable, say whether a journal paper gets accepted or not. The dependent variable or predictors are all the phrases used in these papers - (unigrams, bigrams, trigrams). One of these phrases has a skewed presence in the 'accepted' class. Including this phrase gives me a classifier with a very high accuracy (more than 90%), while removing this phrase results in accuracy dropping to about 70%.
My more general (naive) machine learning question is:
Is it advisable to remove such skewed features when doing classification?
Is there a method to check skewed presence for every feature and then decide whether to keep it in the model or not?
If I understand correctly you ask whether some feature should be removed because it is a good predictor (it makes your classifier works better). So the answer is short and simple - do not remove it in fact, the whole concept is to find exactly such features.
The only reason to remove such feature would be that this phenomena only occurs in the training set, and not in real data. But in such case you have wrong data - which does not represnt the underlying data density and you should gather better data or "clean" the current one so it has analogous characteristics as the "real ones".
Based on your comments, it sounds like the feature in your documents that's highly predictive of the class is a near-tautology: "paper accepted on" correlates with accepted papers because at least some of the papers in your database were scraped from already-accepted papers and have been annotated by the authors as such.
To me, this sounds like a useless feature for trying to predict whether a paper will be accepted, because (I'd imagine) you're trying to predict paper acceptance before the actual acceptance has been issued ! In such a case, none of the papers you'd like to test your algorithm with will be annotated with "paper accepted on." So, I'd remove it.
You also asked about how to determine whether a feature correlates strongly with one class. There are three things that come to mind for this problem.
First, you could just compute a basic frequency count for each feature in your dataset and compare those values across classes. This is probably not super informative, but it's easy.
Second, since you're using a log-linear model, you can train your model on your training dataset, and then rank each feature in your model by its weight in the logistic regression parameter vector. Features with high positive weight are indicative of one class, while features with large negative weight are strongly indicative of the other.
Finally, just for the sake of completeness, I'll point out that you might also want to look into feature selection. There are many ways of selecting relevant features for a machine learning algorithm, but I think one of the most intuitive from your perspective might be greedy feature elimination. In such an approach, you train a classifier using all N features in your model, and measure the accuracy on some held-out validation set. Then, train N new models, each with N-1 features, such that each model eliminates one of the N features, and measure the resulting drop in accuracy. The feature with the biggest drop was probably strongly predictive of the class, while features that have no measurable difference can probably be omitted from your final model. As larsmans points out correctly in the comments below, this doesn't scale well at all, but it can be a useful method sometimes.
I am making a document classifier in mahout using the simple naive bayes algorithm. Currently, 98% of the data(documents) I have is of Class A and only 2% is of class B. My question is, since there is such a wide gap in the percentage of Class A docs vs Class B docs, would the classifier be able to train accurately still?
What I'm thinking of doing is ignoring a whole bunch of Class A documents and "manipulating" the dataset I have so that there isn't such a wide gap in the composition of the documents. Thus, the dataset I'll end up having will consist 30% of Class B and 70% of Class A. But, are there any repercussions of doing that I am not aware of?
A lot of this gets into how good "accuracy" is as a measure of performance, and that depends on your problem. If misclassifying "A" as "B" is just as bad/ok as misclassifying "B" as "A", then there is little reason to do anything other than just mark everything as "A", since you know it will reliably get you a 98% accuracy (so long as that unbalanced distribution is representative of the true distribution).
Without knowing your problem (and if accuracy is the measure you should use), the best answer I could give is "it depends on the data set". It is possible that you could get past 99% accuracy with standard naive bays, though it may be unlikely. For Naive Bayes in particular, one thing you could do is to disable the use of priors (the prior is essentially the proportion of each class). This has the effect of pretending that every class is equally likely to occur, though the model parameters will have been learned from uneven amounts of data.
Your proposed solution is a common practice, it sometimes works well. Another practice is to create fake data for the smaller class (how would depend on your data, for text documents I'm not aware of any particularly good way). Another practice is to increase the weights of the data points in the under-represented classes.
You can search for "imbalanced classification" and find a lot more information about these types of problems (they are one of the harder ones).
If accuracy is not actually a good measure for your problem, you can search for more information about "cost sensitive classification" which should be helpful.
You should not necessarily sample dataset A to reduce its instances. Several methods are available for efficient learning from imbalanced datasets, such as Majority Undersampling (exactly what you did), Minority Oversampling, SMOTE, and etc. Here is an empirical comparison of these methods: http://machinelearning.org/proceedings/icml2007/papers/62.pdf
Alternatively, you may define a custom cost matrix for the classifier. In other words, assuming B=Positive class, you may define cost(False Positive) < cost(False Negative). In this case, the classifier's output will bias towards the positive class. Here is a very helpful tutorial: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.164.4418&rep=rep1&type=pdf