I'm trying to run Word2Vec first on a very tiny toy dataset that I made up by hand -- just to convince myself I'm doing it correctly before I go for my main dataset. But despite doing 99000 iterations the results weren't great. (Tiger and Lion didn't have as high a similarity as I thought they would).
Toy dataset:
s= [['Tiger', 'Zebra'], ['Tiger', 'Lion', 'Cheetah'],
['Orangutan', 'Bonobo', 'Orangutan', 'Chimpanzee'],
['Dog', 'Cat', 'Mouse'], ['Tiger', 'Rhino'],
['House', 'Car'], ['Antelope', 'Gazelle'],
['Zebra', 'Horse'], ['Tiger', 'Lion', 'Leopard'],
['Cat', 'Mouse'], ['Mouse', 'Hampster', 'Gerbil'],
['Rhino', 'Zebra'], ['Zebra', 'Antelope'],
['Tiger', 'Lion'], ['Lion', 'Tiger', 'Giraffe'],
['Leopard', 'Lion'], ['Leopard', 'Tiger', 'Lion'],
['Tiger', 'Lion'], ['Tiger', 'Lion'],
['Car', 'Van'], ['Car', 'Lorry'],
['Car', 'Van'], ['Car', 'Lorry'],
['Car', 'Van'], ['Car', 'Lorry']
]
In theory should I expect a toy dataset like this to show amazing results if I did large amount of iterations?
Here is the code I'm using:
model = gensim.models.Word2Vec(s, min_count=0, iter=iterations,size=100)
Ps. See here for related discussion.
In my experience, Word2Vec does not work well on tiny or contrived datasets. Sometimes, more iterations (or making the model much smaller in size dimensionality) can eke out some hints of meaningfulness – but nothing like the results from real multi-million-word training sets.
The true power of the algorithm relies on the balance-of-influences learned from large, diverse, naturally-varying text examples.
(As your synthetic dataset isn't even comprehensible language, I'm not sure what "amazing results" would be possible – what's the generalizable patterns that these short, repetitive lists-of-animals should be teaching a model?)
With a small data like yours a general model quickly learns the parameters within just 10-20 iterations, doing more iterations would not result in much change in the prediction, if you may do a lot more iterations it may pick up the errors in the data and performance of the model may decrease. So from a small data it can only learn few things, if you train your model on more data your model may perform lot better.
Related
For a project I am working on, which uses annual financial reports data (of multiple categories) from companies which have been successful or gone bust/into liquidation, I previously created a (fairly well performing) model on AWS Sagemaker using a multiple linear regression algorithm (specifically, the AWS stock algorithm for logistic regression/classification problems - the 'Linear Learner' algorithm)
This model just produces a simple "company is in good health" or "company looks like it will go bust" binary prediction, based on one set of annual data fed in; e.g.
query input: {data:[{
"Gross Revenue": -4000,
"Balance Sheet": 10000,
"Creditors": 4000,
"Debts": 1000000
}]}
inference output: "in good health" / "in bad health"
I trained this model by just ignoring what year for each company the values were from and pilling in all of the annual financial reports data (i.e. one years financial data for one company = one input line) for the training, along with the label of "good" or "bad" - a good company was one which has existed for a while, but hasn't gone bust, a bad company is one which was found to have eventually gone bust; e.g.:
label
Gross Revenue
Balance Sheet
Creditors
Debts
good
10000
20000
0
0
bad
0
5
100
10000
bad
20000
0
4
100000000
I hence used these multiple features (gross revenue, balance sheet...) along with the label (good/bad) in my training input, to create my first model.
I would like to use the same features as before as input (gross revenue, balance sheet..) but over multiple years; e.g take the values from 2020 & 2019 and use these (along with the eventual company status of "good" or "bad") as the singular input for my new model. However I'm unsure of the following:
is this an inappropriate use of logistic regression Machine learning? i.e. is there a more suitable algorithm I should consider?
is it fine, or terribly wrong to try and just use the same technique as before, but combine the data for both years into one input line like:
label
Gross Revenue(2019)
Balance Sheet(2019)
Creditors(2019)
Debts(2019)
Gross Revenue(2020)
Balance Sheet(2020)
Creditors(2020)
Debts(2020)
good
10000
20000
0
0
30000
10000
40
500
bad
100
50
200
50000
100
5
100
10000
bad
5000
0
2000
800000
2000
0
4
100000000
I would personally expect that a company which has gotten worse over time (i.e. companies finances are worse in 2020 than in 2019) should be more likely to be found to be a "bad"/likely to go bust, so I would hope that, if I feed in data like in the above example (i.e. earlier years data comes before later years data, on an input line) my training job ends up creating a model which gives greater weighting to the earlier years data, when making predictions
Any advice or tips would be greatly appreciated - I'm pretty new to machine learning and would like to learn more
UPDATE:
Using Long-Short-Term-Memory Recurrent Neural Networks (LSTM RNN) is one potential route I think I could try taking, but this seems to commonly just be used with multivariate data over many dates; my data only has 2 or 3 dates worth of multivariate data, per company. I would want to try using the data I have for all the companies, over the few dates worth of data there are, in training
I once developed a so called Genetic Time Series in R. I used a Genetic Algorithm which sorted out the best solutions from multivariate data, which were fitted on a VAR in differences or a VECM. Your data seems more macro economic or financial than user-centric and VAR or VECM seems appropriate. (Surely it is possible to treat time-series data in the same way so that we can use LSTM or other approaches, but these are very common) However, I do not know if VAR in differences or VECM works with binary classified labels. Perhaps if you would calculate a metric outcome, which you later label encode to a categorical feature (or label it first to a categorical) than VAR or VECM may also be appropriate.
However you may add all yearly data points to one data points per firm to forecast its survival, but you would loose a lot of insight. If you are interested in time series ML which works a little bit different than for neural networks or elastic net (which could also be used with time series) let me know. And we can work something out. Or I'll paste you some sources.
Summary:
1.)
It is possible to use LSTM, elastic NEt (time points may be dummies or treated as cross sectional panel) or you use VAR in differences and VECM with a slightly different out come variable
2.)
It is possible but you will loose information over time.
All the best,
Patrick
I am interested in the field of topic modelling and I am about to develop my own algorithm. The problem I am facing at the moment is how to compare the results of my work with the results of state-of-the art models such as LDA. Since the results of the LDA are non-deterministic (to the best of my knowledge they can not be set to be the same if the algorithm is run a second time), how could I draw a conclusion about how good my model is compared to the LDA model using different configuration sets?
Run several times, and report the best and average results.
That is the standard approach to handle random starting conditions.
As long as the spread of the results is not too large, this will still support some conclusions.
I have moderate experience with data science. I have a data set with 9500 observations and more than 4500 features most of which are highly correlated. Here is briefly what I have tried: I have dropped columns where there are less than 6000 non-NAs and have imputed NAs with their corresponding columns' median values when there are at least 6000 non-NAs. As for correlation, I have kept only features having at most 0.7 correlation with others. By doing so, I have reduced the number of features to about 750. Then I have used those features in my binary classification task in random forest.
My data set is highly unbalanced where ratio of (0:1) is (10:1). So when I apply RF with 10-fold cv, I observe too good results in each cv (AUC of 99%) which is to good to be true and in my test set I got way worse results such as 0.7. Here is my code:
import h2o
from h2o.estimators import H2ORandomForestEstimator
h2o.init(port=23, nthreads=4)
train = fs_rf[fs_rf['Year'] <= '201705']
test = fs_rf[fs_rf['Year'] > '201705']
train = train.drop('Year',axis=1)
test = test.drop('Year',axis=1)
test.head()
train = h2o.H2OFrame(train)
train['BestWorst2'] = train['BestWorst2'].asfactor()
test = h2o.H2OFrame(test)
test['BestWorst2'] = test['BestWorst2'].asfactor()
training_columns = train.drop('BestWorst2',axis=1).col_names
response_column = 'BestWorst2'
model = H2ORandomForestEstimator(ntrees=100, max_depth=20, nfolds=10, balance_classes=True)
model.train(x=training_columns, y=response_column, training_frame=train)
performance = model.model_performance(test_data=test)
print(performance)
How could I avoid this over-fitting? I have tried many different parameters in grid search but none of them improved the results.
This is not what I would call "overfitting". The reason you are seeing really good cross-validation metrics compared to your test metrics is that you have time-series data and so you can't use k-fold cross-validation to give you an accurate estimate of performance.
Performing k-fold cross-validation on a time-series dataset will give you overly-optimistic performance metrics because you are not respecting the time-series component in your data. Regular k-fold cross-validation will randomly sample from your whole dataset to create a train & validation set. Essentially, your validation strategy is "cheating" because you have "future" data included in your CV training sets (if that makes any sense).
I can see by your code that you understand that you need to train with "past" data and predict on "future" data, but if you want to read more about this topic, I'd recommend this article or this article.
One solution is to simply look at test set performance as way to evaluate your model. Another option is to use what's called "rolling" or "time-series" cross-validation, but H2O does not currently support that (though it seems like it might be added soon). Here's a ticket for this if you want to keep track of the progress.
can word2vec be used for guessing words with just context?
having trained the model with a large data set e.g. Google news how can I use word2vec to predict a similar word with only context e.g. with input ", who dominated chess for more than 15 years, will compete against nine top players in St Louis, Missouri." The output should be Kasparov or maybe Carlsen.
I'ven seen only the similarity apis but I can't make sense how to use them for this? is this not how word2vec was intented to use?
It is not the intended use of word2vec. The word2vec algorithm internally tries to predict exact words, using surrounding words, as a roundabout way to learn useful vectors for those surrounding words.
But even so, it's not forming exact predictions during training. It's just looking at a single narrow training example – context words and target word – and performing a very simple comparison and internal nudge to make its conformance to that one example slightly better. Over time, that self-adjusts towards useful vectors – even if the predictions remain of wildly-varying quality.
Most word2vec libraries don't offer a direct interface for showing ranked predictions, given context words. The Python gensim library, for the last few versions (as of current version 2.2.0 in July 2017), has offered a predict_output_word() method that roughly shows what the model would predict, given context-words, for some training modes. See:
https://radimrehurek.com/gensim/models/word2vec.html#gensim.models.word2vec.Word2Vec.predict_output_word
However, considering your fill-in-the-blank query (also called a 'cloze deletion' in related education or machine-learning contexts):
_____, who dominated chess for more than 15 years, will compete against nine top players in St Louis, Missouri
A vanilla word2vec model is unlikely to get that right. It has little sense of the relative importance of words (except when some words are more narrowly predictive of others). It has no sense of grammar/ordering, or or of the compositional-meaning of connected-phrases (like 'dominated chess' as opposed to the separate words 'dominated' and 'chess'). Even though words describing the same sorts of things are usually near each other, it doesn't know categories to be able to determine that the blank must be a 'person' and a 'chess player', and the fuzzy-similarities of word2vec don't guarantee words-of-a-class will necessarily all be nearer-each-other than other words.
There has been a bunch of work to train word/concept vectors (aka 'dense embeddings') to be better at helping at such question-answering tasks. A random example might be "Creating Causal Embeddings for Question Answering with Minimal Supervision" but queries like [word2vec question answering] or [embeddings for question answering] will find lots more. I don't know of easy out-of-the-box libraries for doing this, with or without a core of word2vec, though.
I am trying to predict whether a particular service ticket raised by client needs a code change.
I have training data.
I have around 17k data points with problem description and tag (Y for code change required and N for no code change)
I did TF-IDF and it gave me 27k features. So I tried to fit RandomForestClassifier (sklearn python) with this 17k x 27k matrix.
I am getting very low scores on test set while training accuracy is very high.
Precision on train set: 89%
Precision on test set: 21%
Can someone suggest any workarounds?
I am using this model now:
sklearn.RandomForestClassifier(n_jobs=3,n_estimators=100,class_weight='balanced',max_features=None,oob_score=True)
Please help!
EDIT:
I have 11k training data with 900 positives (skewed). I tried LinearSVC sparsify but didn't work as well as Truncated SVD (Latent Semantic Indexing). maxFeatures=None performs better on the test set than without it.
I have also tried SVM, logistic (l2 and l1), ExtraTrees. RandonForest still is working best.
Right now, going at 92% precision on positives but recall is 3% only
Any other suggestions would be appreciated!
Update:
Feature engineering helped a lot. I pulled features out of the air (len of chars, len of words, their, difference, ratio, day of week the problem was of reported, day of month, etc) and now I am at 19-20% recall with >95% accuracy.
Food for your thoughts on using word2vec average vectors as deep features for the free text instead of tf-idf or bag of words ???
[edited]
Random forest handles more features than data points quite fine. RF is e.g. used for micro-array studies with e.g. a 100:5000 data point/feature ratio or in single-nucleotide_polymorphism(SNP) studies with e.g 5000:500,000 ratio.
I do disagree with the diagnose provided by #ncfirth, but the suggested treatment of variable selection may help anyway.
Your default random forest is not badly overfitted. It is just not meaningful to pay any attention to a non-cross validated training set prediction performance for a RF model, because any sample will end in the terminal nodes/leafs it has itself defined. But the overall ensemble model is still robust.
[edit] If you would change the max_depth or min_samples_split, the training precision would probably drop, but that is not the point. The non-cross validated training error/precision of a random forest model or many other ensemble models simply does not estimate anything useful.
[I did before edit confuse max_features with n_estimators, sry I mostly use R]
Setting max_features="none" is not random forest, but rather 'bagged trees'. You may benefit from a somewhat lower max_features which improve regularization and speed, maybe not. I would try lowering max_features to somewhere between 27000/3 and sqrt(27000), the typical optimal range.
You may achieve better test set prediction performance by feature selection. You can run one RF model, keep the top ~5-50% most important features and then re-run the model with fewer features. "L1 lasso" variable selection as ncfirth suggests may also be a viable solution.
Your metric of prediction performance, precision, may not be optimal in case unbalanced data or if the cost of false-negative and false-positive is quite different.
If your test set is still predicted much worse than the out-of-bag cross-validated training set, you may have problems with your I.I.D. assumptions that any supervised ML model rely on or you may need to wrap the entire data processing in an outer cross-validation loop, to avoid over optimistic estimation of prediction performance due to e.g. the variable selection step.
Seems like you've overfit on your training set. Basically the model has learnt noise on the data rather than the signal. There are a few ways to combat this, but it seems fairly obvious that you're model has overfit because of the incredibly large number of features you're feeding it.
EDIT:
It seems I was perhaps too quick to jump to the conclusion of overfitting, however this may still be the case (left as an exercise to the reader!). However feature selection may still improve the generalisability and reliability of your model.
A good place to start for removing features in scikit-learn would be here. Using sparsity is a fairly common way to perform feature selection:
from sklearn.svm import LinearSVC
from sklearn.feature_selection import SelectFromModel
import numpy as np
# Create some data
X = np.random.random((1800, 2700))
# Boolean labels as the y vector
y = np.random.random(1800)
y = y > 0.5
y = y.astype(bool)
lsvc = LinearSVC(C=0.05, penalty="l1", dual=False).fit(X, y)
model = SelectFromModel(lsvc, prefit=True)
X_new = model.transform(X)
print X_new.shape
Which returns a new matrix of shape (1800, 640). You can tune the number of features selected by altering the C parameter (called the penalty parameter in scikit-learn but sometimes called the sparsity parameter).