I'm working on text classification and I have a set of 200.000 tweets.
The idea is to manually label a short set of tweets and train classifiers to predict the labels of the rest. Supervised learning.
What I would like to know is if there is a method to choose what samples to include in the train set in a way that this train set is a good representation of the whole data set, and because the high diversity included in the train set, the trained classifiers have considerable trust to be applied on the rest of tweets.
This sounds like a stratification question - do you have pre-existing labels or do you plan to design the labels based on the sample you're constructing?
If it's the first scenario, I think the steps in order of importance would be:
Stratify by target class proportions (so if you have three classes, and they are 50-30-20%, train/dev/test should follow the same proportions)
Stratify by features you plan to use
Stratify by tweet length/vocabulary etc.
If it's the second scenario, and you don't have labels yet, you may want to look into using n-grams as a feature, coupled with a dimensionality reduction or clustering approach. For example:
Use something like PCA or t-SNE to maximize distance between tweets (or a large subset), then pick candidates from different regions of the projected space
Cluster them based on lexical items (unigrams or bigrams, possibly using log frequencies or TF-IDF and stop word filtering, if content words are what you're looking for) - then you can cut the tree at a height that gives you n bins, which you can then use as a source for samples (stratify by branch)
Use something like LDA to find n topics, then sample stratified by topic
Hope this helps!
It seems that before you know anything about the classes you are going to label, a simple uniform random sample will do almost as well as any stratified sample - because you don't know in advance what to stratify on.
After labelling this first sample and building the first classifier, you can start so-called active learning: make predictions for the unlabelled dataset, and sample some tweets in which your classifier is least condfident. Label them, retrain the classifier, and repeat.
Using this approach, I managed to create a good training set after several (~5) iterations, with ~100 texts in each iteration.
Related
I need to classify website text with zero or more categories/labels (5 labels such as finance, tech, etc). My problem is handling text that isn't one of these labels.
I tried ML libraries (maxent, naive bayes), but they match "other" text incorrectly with one of the labels. How do I train a model to handle the "other" text? The "other" label is so broad and it's not possible to pick a representative sample.
Since I have no ML background and don't have much time to build a good training set, I'd prefer a simpler approach like a term frequency count, using a predefined list of terms to match for each label. But with the counts, how do I determine a relevancy score, i.e. if the text is actually that label? I don't have a corpus and can't use tf-idf, etc.
Another idea , is to user neural networks with softmax output function, softmax will give you a probability for every class, when the network is very confident about a class, will give it a high probability, and lower probabilities to the other classes, but if its insecure, the differences between probabilities will be low and none of them will be very high, what if you define a treshold like : if the probability for every class is less than 70% , predict "other"
Whew! Classic ML algorithms don't combine both multi-classification and "in/out" at the same time. Perhaps what you could do would be to train five models, one for each class, with a one-against-the-world training. Then use an uber-model to look for any of those five claiming the input; if none claim it, it's "other".
Another possibility is to reverse the order of evaluation: train one model as a binary classifier on your entire data set. Train a second one as a 5-class SVM (for instance) within those five. The first model finds "other"; everything else gets passed to the second.
What about creating histograms? You could use a bag of words approach using significant indicators of for e.g. Tech and Finance. So, you could try to identify such indicators by analyzing the certain website's tags and articles or just browse the web for such inidicators:
http://finance.yahoo.com/news/most-common-words-tech-finance-205911943.html
Let's say your input vactor X has n dimensions where n represents the number of indicators. For example Xi then holds the count for the occurence of the word "asset" and Xi+k the count of the word "big data" in the current article.
Instead of defining 5 labels, define 6. Your last category would be something like a "catch-all" category. That's actually your zero-match category.
If you must match the zero or more category, train a model which returns probability scores (such as a neural net as Luis Leal suggested) per label/class. You could than rate your output by that score and say that every class with a score higher than some threshold t is a matching category.
Try this NBayes implementation.
For identifying "Other" categories, dont bother much. Just train on your required categories which clearly identifies them, and introduce a threshold in the classifier.
If the values for a label does not cross a threshold, then the classifier adds the "Other" label.
It's all in the training data.
AWS Elasticsearch percolate would be ideal, but we can't use it due to the HTTP overhead of percolating documents individually.
Classify4J appears to be the best solution for our needs because the model looks easy to train and it doesn't require training of non-matches.
http://classifier4j.sourceforge.net/usage.html
I'm writing a naive bayes classifier for a class project and I just got it working... sort of. While I do get an error-free output, the winning output label had an output probability of 3.89*10^-85.
Wow.
I have a couple of ideas of what I might be doing wrong. Firstly, I am not normalizing the output percentages for the classes, so all of the percentages are effectively zero. While that would give me numbers that look nice, I don't know if that's the correct thing to do.
My second idea was to reduce the number of features. Our input data is a list of pseudo-images in the form of a very long text file. Currently, our features are just the binary value of every pixel of the image, and with a 28x28 image that's a lot of features. If I instead chopped the image into blocks of size, say, 7x7, how much would that actually improve the output percentages?
tl;dr Here's the general things I'm trying to understand about naive bayes:
1) Do you need to normalize the output percentages from testing each class?
2) How much of an effect does having too many features have on the results?
Thanks in advance for any help you can give me.
It could be normal. The output of a naive bayes is not meant to be a real probability. What it is meant to do is order a score among competing classes.
The reason why the probability is so low is that many Naive Bayes implementations are the product of the probabilities of all the observed features of the instance that is being classified. If you are classifying text, each feature may have a low conditional probability for each class (example: lower than 0.01). If you multiply 1000s of feature probabilities, you quickly end up with numbers such as you have reported.
Also, the probabilities returned are not the probabilities of each class given the instance, but an estimate of the probabilities of observing this set of features, given the class. Thus, the more you have features, the less likely it is to observe these exact features. A bayesian theorem is used to change argmax_c P(class_c|features) to argmax_c P(class_c)*P(features|class_c), and then the P(features|class_c) is further simplified by making independence assumption, which allows changing that to a product of the probabilities of observing each individual feature given the class. These assumptions don't change the argmax (the winning class).
If I were you, I would not really care about the probability output, focus instead on the accuracy of your classifier and take action to improve the accuracy, not the calculated probabilities.
My problem is that I have a large unlabeled dataset, but over time I want it to become labeled and build a confident classifier.
This can be done by active learning, but active learning needs an initial classifier to be built for it to then estimate and rank the remaining unlabeled instances by how informative they are expected to be to the classifier.
To build the initial classifier, I need to label some examples by hand. my questions is: Are there methods to find likely informative examples in the initial unlabeled dataset, without the help of an initial classifier?
I thought about just using k-means with some number of clusters, run it and label one example from each cluster, then train the classifier on these.
Is there a better way?
I have to disagree with Edward Raff.
k-means may turn out to be useful here (if your data is continuous).
Just use a rather large value of k.
The idea is to avoid picking too similar objects, but get a sample that covers the data reasonably well. k-means may fail to "cluster" complex data, but it works reasonably well for quantization. So it will return a "less random, more representative" sample from your data.
But beware: k-means centers do not correspond to data points. You could either use a medoid based algorithm, or just find the closes instance to each center.
Some alternatives:
if you can afford to label "a" objects, run k-means with k=a
run k-means with k=5*a, and select 20% of the centers (maybe preferring those with highest density)
choose 0.5*a by k-means, 0.5*a randomly
do either, but choose only 0.5*a objects to label. Train a classifier, find the 0.5*a unlabeled objects that the classifier had the lowest confidence on
No. If you don't have any labeled data, you have no way of determining which points are the most informative. k-means does not necessarily help either, as you don't know where the decision surface lives.
You are overthinking the problem. Just randomly sample some data and get it labeled. Once you have a few hundred - thousand points labeled you can start to look at the labeled data and makes some decisions about where to head next.
I've got a problem where I've potentially got a huge number of features. Essentially a mountain of data points (for discussion let's say it's in the millions of features). I don't know what data points are useful and what are irrelevant to a given outcome (I guess 1% are relevant and 99% are irrelevant).
I do have the data points and the final outcome (a binary result). I'm interested in reducing the feature set so that I can identify the most useful set of data points to collect to train future classification algorithms.
My current data set is huge, and I can't generate as many training examples with the mountain of data as I could if I were to identify the relevant features, cut down how many data points I collect, and increase the number of training examples. I expect that I would get better classifiers with more training examples given fewer feature data points (while maintaining the relevant ones).
What machine learning algorithms should I focus on to, first,
identify the features that are relevant to the outcome?
From some reading I've done it seems like SVM provides weighting per feature that I can use to identify the most highly scored features. Can anyone confirm this? Expand on the explanation? Or should I be thinking along another line?
Feature weights in a linear model (logistic regression, naive Bayes, etc) can be thought of as measures of importance, provided your features are all on the same scale.
Your model can be combined with a regularizer for learning that penalises certain kinds of feature vectors (essentially folding feature selection into the classification problem). L1 regularized logistic regression sounds like it would be perfect for what you want.
Maybe you can use PCA or Maximum entropy algorithm in order to reduce the data set...
You can go for Chi-Square tests or Entropy depending on your data type. Supervized discretization highly reduces the size of your data in a smart way (take a look into Recursive Minimal Entropy Partitioning algorithm proposed by Fayyad & Irani).
If you work in R, the SIS package has a function that will do this for you.
If you want to do things the hard way, what you want to do is feature screening, a massive preliminary dimension reduction before you do feature selection and model selection from a sane-sized set of features. Figuring out what is the sane-size can be tricky, and I don't have a magic answer for that, but you can prioritize what order you'd want to include the features by
1) for each feature, split the data in two groups by the binary response
2) find the Komogorov-Smirnov statistic comparing the two sets
The features with the highest KS statistic are most useful in modeling.
There's a paper "out there" titled "A selctive overview of feature screening for ultrahigh-dimensional data" by Liu, Zhong, and Li, I'm sure a free copy is floating around the web somewhere.
4 years later I'm now halfway through a PhD in this field and I want to add that the definition of a feature is not always simple. In the case that your features are a single column in your dataset, the answers here apply quite well.
However, take the case of an image being processed by a convolutional neural network, for example, a feature is not one pixel of the input, rather it's much more conceptual than that. Here's a nice discussion for the case of images:
https://medium.com/#ageitgey/machine-learning-is-fun-part-3-deep-learning-and-convolutional-neural-networks-f40359318721
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.