I'm working with sentiment analysis using NB classifier. I've found some information (blogs, tutorials etc) that training corpus should be balanced:
33.3% Positive;
33.3% Neutral
33.3% Negative
My question is:
Why corspus should be balanced? The Bayes theorem is based on propability of reason/case. So for training purpose isn't it important that in real world for example negative tweets are only 10% not 33.3%?
You are correct, balancing data is important for many discriminative models, but not really for NB.
However, it might be still more beneficial to bias P(y) estimators to get better predictive performance (since due to various simplifications models use, probability assigned to minority class can be heaviy underfitted). For NB it is not about balancing data, but literally modifying the estimated P(y) so that on the validation set accuracy is maximised.
In my opinion the best dataset for training purposes if a sample of the real world data that your classifier will be used with.
This is true for all classifiers (but some of them are indeed not suitable to unbalanced training sets in which cases you don't really have a choice to skew the distribution), but particularly for probabilistic classifiers such as Naive Bayes. So the best sample should reflect the natural class distribution.
Note that this is important not only for the class priors estimates. Naive Bayes will calculate for each feature the likelihood of predicting the class given the feature. If your bayesian classifier is built specifically to classify texts, it will use global document frequency measures (the number of times a given word occurs in the dataset, across all categories). If the number of documents per category in the training set doesn't reflect their natural distribution, the global term frequency of terms usually seen in unfrequent categories will be overestimated, and that of frequent categories underestimated. Thus not only the prior class probability will be incorrect, but also all the P(category=c|term=t) estimates.
Related
I am interested in any tips on how to train a set with a very limited positive set and a large negative set.
I have about 40 positive examples (quite lengthy articles about a particular topic), and about 19,000 negative samples (most drawn from the sci-kit learn newsgroups dataset). I also have about 1,000,000 tweets that I could work with.. negative about the topic I am trying to train on. Is the size of the negative set versus the positive going to negatively influence training a classifier?
I would like to use cross-validation in sci-kit learn. Do I need to break this into train / test-dev / test sets? Is know there are some pre-built libraries in sci-kit. Any implementation examples that you recommend or have used previously would be helpful.
Thanks!
The answer to your first question is yes, the amount by which it will affect your results depends on the algorithm. My advive would be to keep an eye on the class-based statistics such as recall and precision (found in classification_report).
For RandomForest() you can look at this thread which discusses
the sample weight parameter. In general sample_weight is what
you're looking for in scikit-learn.
For SVM's have a look at either this example or this
example.
For NB classifiers, this should be handled implicitly by Bayes
rule, however in practice you may see some poor performances.
For you second question it's up for discussion, personally I break my data into a training and test split, perform cross validation on the training set for parameter estimation, retrain on all the training data and then test on my test set. However the amount of data you have may influence the way you split your data (more data means more options).
You could probably use Random Forest for your classification problem. There are basically 3 parameters to deal with data imbalance. Class Weight, Samplesize and Cutoff.
Class Weight-The higher the weight a class is given, the more its error rate is decreased.
Samplesize- Oversample the minority class to improve class imbalance while sampling the defects for each tree[not sure if Sci-kit supports this, used to be param in R)
Cutoff- If >x% trees vote for the minority class, classify it as minority class. By default x is 1/2 in Random forest for 2-class problem. You can set it to a lower value for the minority class.
Check out balancing predict error at https://www.stat.berkeley.edu/~breiman/RandomForests/cc_home.htm
For the 2nd question if you are using Random Forest, you do not need to keep separate train/validation/test set. Random Forest does not choose any parameters based on a validation set, so validation set is un-necessary.
Also during the training of Random Forest, the data for training each individual tree is obtained by sampling by replacement from the training data, thus each training sample is not used for roughly 1/3 of the trees. We can use the votes of these 1/3 trees to predict the out of box probability of the Random forest classification. Thus with OOB accuracy you just need a training set, and not validation or test data to predict performance on unseen data. Check Out of Bag error at https://www.stat.berkeley.edu/~breiman/RandomForests/cc_home.htm for further study.
I have been trying to figure out the correlation between the error rate and the number of features in both of these models. I watched some videos, and the creator of the video said that a simple model can be better than a complicated model. So I figured that the more features I had the greater the error rate would be. This did not prove to be true in my work, and when I had less features the error rate went up. I'm not sure if I'm doing this incorrectly, or if the guy in the video made a mistake. Can someone care to explain? I also am curious how features relate to Logistic Regression's error rate as well.
Naive Bayes and Logistic Regression are a "generative-discriminative pair," meaning they have the same model form (a linear classifier), but they estimate parameters in different ways.
For feature x and label y, naive Bayes estimates a joint probability p(x,y) = p(y)*p(x|y) from the training data (that is, builds a model that could "generate" the data), and uses Bayes Rule to predict p(y|x) for new test instances. On the other hand, logistic regression estimates p(y|x) directly from the training data by minimizing an error function (which is more "discrimative").
These differences have implications for error rate:
When there are very few training instances, logistic regression might "overfit," because there isn't enough data to estimate p(y|x) reliably. Naive Bayes might do better because it models the entire joint distribution.
When the feature set is large (and sparse, like word features in text classification) naive Bayes might "double count" features that are correlated with each other, because it assumes that each p(x|y) event is independent, when they are not. Logistic regression can do a better job by naturally "splitting the difference" among these correlated features.
If the features really are (mostly) conditionally independent, both models might actually improve with more and more features, provided there are enough data instances. The problem comes when the training set size is small relative to the number of features. Priors on naive Bayes feature parameters, or regularization methods (like L1/Lasso or L2/Ridge) on logistic regression can help in these cases.
In my bachelor thesis I am supposed to use AdaBoostM1 with a MultinomialNaiveBayes classifier on a text classification problem. The problem is that in most cases, the M1 is worse or equal to the MultinomialNaiveBayes without boosting.
I use the following code:
AdaBoostM1 m1 = new AdaBoostM1();
m1.setClassifier(new NaiveBayesMultinomial());
m1.buildClassifier(training);
So I don't get how the AdaBoost would not be able to improve the results? Unfortunately, I couldn't find anything else about that on the web as most people seem to be very satisfied with the AdaBoost.
AdaBoost is a binary/dichotomous/2-class classifier and designed to boost a weak learner that is just better than 1/2 accuracy. AdaBoostM1 is a M-class classifier but still requires the weak learner to be better than 1/2 accuracy, when one would expect chance level to be around 1/M. Balancing/weighting is used to get equal prevalence classes initially, but the reweighting inherent to AdaBoost can destroy this quickly. A solution is to base boosting on chance corrected measures like Kappa or Informedness (AdaBook).
As M grows, e.g. with text classification, this mismatch grows, and thus a much stronger than chance classifier is needed. Thus with M=100, chance is about 1% but 50% minimum accuracy is needed by AdaBoostM1.
As base classifiers get stronger (viz. no longer barely above chance) the scope for boosting to improve things reduces - it has already pulled us to a very specific part of the search space. It is increasingly likely to have overfitted to errors and outliers, so there is no scope to balance a wide variety of variants.
A number of resources on informedness (including matlab code and xls sheets and early papers) are here: http://david.wardpowers.info/BM A comparison with other chance-corrected kappa measures is here: http://aclweb.org/anthology-new/E/E12/E12-1035.pdf
A weka implementation and experimentation for Adaboost using Bookmaker informedness is available - contact author.
It's hard to beat Naive Bayes on text classification. Furthermore, boosting was designed for weak classifiers with high bias and that's where boosting performs well. Boosting decreases bias but increases variance. Hence if you want the combo AdaBoost + Naive Bayes to outperform Naive Bayes you have to have a big training data set and cross the border, where enlarging of the training set doesn't further increase Naive Bayes's performance (while AdaBoost still benefits from the enlarged training data set).
You may want to read the following paper which examines boosting on Naive Bayes. It demonstrates that boosting does not improve the accuracy of the naive Bayesian classifier as much is usually expected in a set of natural domains:
http://onlinelibrary.wiley.com/doi/10.1111/1467-8640.00219/abstract
Hope it provides a good insight.
I should decide between SVM and neural networks for some image processing application. The classifier must be fast enough for near-real-time application and accuracy is important too. Since this is a medical application, it is important that the classifier has the low failure rate.
which one is better choice?
A couple of provisos:
performance of a ML classifier can refer to either (i) performance of the classifier itself; or (ii) performance of the predicate step: execution speed of the model-building algorithm. Particularly in this case, the answer is quite different depending on which of the two is intended in the OP, so i'll answer each separately.
second, by Neural Network, i'll assume you're referring to the most common implementation--i.e., a feed-forward, back-propagating single-hidden-layer perceptron.
Training Time (execution speed of the model builder)
For SVM compared to NN: SVMs are much slower. There is a straightforward reason for this: SVM training requires solving the associated Lagrangian dual (rather than primal) problem. This is a quadratic optimization problem in which the number of variables is very large--i.e., equal to the number of training instances (the 'length' of your data matrix).
In practice, two factors, if present in your scenario, could nullify this advantage:
NN training is trivial to parallelize (via map reduce); parallelizing SVM training is not trivial, but it's also not impossible--within the past eight or so years, several implementations have been published and proven to work (https://bibliographie.uni-tuebingen.de/xmlui/bitstream/handle/10900/49015/pdf/tech_21.pdf)
mult-class classification problem SVMs are two-class classifiers.They can be adapted for multi-class problems, but this is never straightforward because SVMs use direct decision functions. (An excellent source for modifying SVMs to multi-class problems is S. Abe, Support Vector Machines for Pattern Classification, Springer, 2005). This modification could wipe out any performance advantage SVMs have over NNs: So for instance, if your data has
more than two classes and you chose to configure the SVM using
successive classificstaion (aka one-against-many classification) in
which data is fed to a first SVM classifier which classifiers the
data point either class I or other; if the class is other then
the data point is fed to a second classifier which classifies it
class II or other, etc.
Prediction Performance (execution speed of the model)
Performance of an SVM is substantially higher compared to NN. For a three-layer (one hidden-layer) NN, prediction requires successive multiplication of an input vector by two 2D matrices (the weight matrices). For SVM, classification involves determining on which side of the decision boundary a given point lies, in other words a cosine product.
Prediction Accuracy
By "failure rate" i assume you mean error rate rather than failure of the classifier in production use. If the latter, then there is very little if any difference between SVM and NN--both models are generally numerically stable.
Comparing prediction accuracy of the two models, and assuming both are competently configured and trained, the SVM will outperform the NN.
The superior resolution of SVM versus NN is well documented in the scientific literature. It is true that such a comparison depends on the data, the configuration, and parameter choice of the two models. In fact, this comparison has been so widely studied--over perhaps all conceivable parameter space--and the results so consistent, that even the existence of a few exceptions (though i'm not aware of any) under impractical circumstances shouldn't interfere with the conclusion that SVMs outperform NNs.
Why does SVM outperform NN?
These two models are based on fundamentally different learing strategies.
In NN, network weights (the NN's fitting parameters, adjusted during training) are adjusted such that the sum-of-square error between the network output and the actual value (target) is minimized.
Training an SVM, by contrast, means an explicit determination of the decision boundaries directly from the training data. This is of course required as the predicate step to the optimization problem required to build an SVM model: minimizing the aggregate distance between the maximum-margin hyperplane and the support vectors.
In practice though it is harder to configure the algorithm to train an SVM. The reason is due to the large (compared to NN) number of parameters required for configuration:
choice of kernel
selection of kernel parameters
selection of the value of the margin parameter
I am using a Naive Bayes Classifier to categorize several thousand documents into 30 different categories. I have implemented a Naive Bayes Classifier, and with some feature selection (mostly filtering useless words), I've gotten about a 30% test accuracy, with 45% training accuracy. This is significantly better than random, but I want it to be better.
I've tried implementing AdaBoost with NB, but it does not appear to give appreciably better results (the literature seems split on this, some papers say AdaBoost with NB doesn't give better results, others do). Do you know of any other extensions to NB that may possibly give better accuracy?
In my experience, properly trained Naive Bayes classifiers are usually astonishingly accurate (and very fast to train--noticeably faster than any classifier-builder i have everused).
so when you want to improve classifier prediction, you can look in several places:
tune your classifier (adjusting the classifier's tunable paramaters);
apply some sort of classifier combination technique (eg,
ensembling, boosting, bagging); or you can
look at the data fed to the classifier--either add more data,
improve your basic parsing, or refine the features you select from
the data.
w/r/t naive Bayesian classifiers, parameter tuning is limited; i recommend to focus on your data--ie, the quality of your pre-processing and the feature selection.
I. Data Parsing (pre-processing)
i assume your raw data is something like a string of raw text for each data point, which by a series of processing steps you transform each string into a structured vector (1D array) for each data point such that each offset corresponds to one feature (usually a word) and the value in that offset corresponds to frequency.
stemming: either manually or by using a stemming library? the popular open-source ones are Porter, Lancaster, and Snowball. So for
instance, if you have the terms programmer, program, progamming,
programmed in a given data point, a stemmer will reduce them to a
single stem (probably program) so your term vector for that data
point will have a value of 4 for the feature program, which is
probably what you want.
synonym finding: same idea as stemming--fold related words into a single word; so a synonym finder can identify developer, programmer,
coder, and software engineer and roll them into a single term
neutral words: words with similar frequencies across classes make poor features
II. Feature Selection
consider a prototypical use case for NBCs: filtering spam; you can quickly see how it fails and just as quickly you can see how to improve it. For instance, above-average spam filters have nuanced features like: frequency of words in all caps, frequency of words in title, and the occurrence of exclamation point in the title. In addition, the best features are often not single words but e.g., pairs of words, or larger word groups.
III. Specific Classifier Optimizations
Instead of 30 classes use a 'one-against-many' scheme--in other words, you begin with a two-class classifier (Class A and 'all else') then the results in the 'all else' class are returned to the algorithm for classification into Class B and 'all else', etc.
The Fisher Method (probably the most common way to optimize a Naive Bayes classifier.) To me,
i think of Fisher as normalizing (more correctly, standardizing) the input probabilities An NBC uses the feature probabilities to construct a 'whole-document' probability. The Fisher Method calculates the probability of a category for each feature of the document then combines these feature probabilities and compares that combined probability with the probability of a random set of features.
I would suggest using a SGDClassifier as in this and tune it in terms of regularization strength.
Also try to tune the formula in TFIDF you're using by tuning the parameters of TFIFVectorizer.
I usually see that for text classification problems SVM or Logistic Regressioin when trained one-versus-all outperforms NB. As you can see in this nice article by Stanford people for longer documents SVM outperforms NB. The code for the paper which uses a combination of SVM and NB (NBSVM) is here.
Second, tune your TFIDF formula (e.g. sublinear tf, smooth_idf).
Normalize your samples with l2 or l1 normalization (default in Tfidfvectorization) because it compensates for different document lengths.
Multilayer Perceptron, usually gets better results than NB or SVM because of the non-linearity introduced which is inherent to many text classification problems. I have implemented a highly parallel one using Theano/Lasagne which is easy to use and downloadable here.
Try to tune your l1/l2/elasticnet regularization. It makes a huge difference in SGDClassifier/SVM/Logistic Regression.
Try to use n-grams which is configurable in tfidfvectorizer.
If your documents have structure (e.g. have titles) consider using different features for different parts. For example add title_word1 to your document if word1 happens in the title of the document.
Consider using the length of the document as a feature (e.g. number of words or characters).
Consider using meta information about the document (e.g. time of creation, author name, url of the document, etc.).
Recently Facebook published their FastText classification code which performs very well across many tasks, be sure to try it.
Using Laplacian Correction along with AdaBoost.
In AdaBoost, first a weight is assigned to each data tuple in the training dataset. The intial weights are set using the init_weights method, which initializes each weight to be 1/d, where d is the size of the training data set.
Then, a generate_classifiers method is called, which runs k times, creating k instances of the Naïve Bayes classifier. These classifiers are then weighted, and the test data is run on each classifier. The sum of the weighted "votes" of the classifiers constitutes the final classification.
Improves Naive Bayes classifier for general cases
Take the logarithm of your probabilities as input features
We change the probability space to log probability space since we calculate the probability by multiplying probabilities and the result will be very small. when we change to log probability features, we can tackle the under-runs problem.
Remove correlated features.
Naive Byes works based on the assumption of independence when we have a correlation between features which means one feature depends on others then our assumption will fail.
More about correlation can be found here
Work with enough data not the huge data
naive Bayes require less data than logistic regression since it only needs data to understand the probabilistic relationship of each attribute in isolation with the output variable, not the interactions.
Check zero frequency error
If the test data set has zero frequency issue, apply smoothing techniques “Laplace Correction” to predict the class of test data set.
More than this is well described in the following posts
Please refer below posts.
machinelearningmastery site post
Analyticvidhya site post
keeping the n size small also make NB to give high accuracy result. and at the core, as the n size increase its accuracy degrade,
Select features which have less correlation between them. And try using different combination of features at a time.