I have a question about variable importance ranking.
I built an MLP and an RF model using the same dataset with 34 variables and achieved the same accuracy on a similar test dataset. As you can see in the picture below the top variables for the SHAP summary plot and the RF VIM are quite different.
Interestingly, I removed the low-ranked variable from the MLP and the accuracy increased. However, the RF result didn’t change.
Does that mean the RF is not a good choice for modeling this dataset?
It’s still strange to me that the rankings are so different:
SHAP summary plot vs. RF VIM, I numbered the top and low-ranked variable
Shouldn't the variables ranking be the same for MLP and RF?
No. There may be tendency for different algos to rank certain features higher, but there is no reason for ranking to be the same.
Different algorithms:
May have different objective functions to achieve intended goal.
May use features differently to achieve min (max) of the objective function.
On top, what you cite as RF "feature importances" (mean decrease in Gini) is only one of the many ways to calculate "feature importance" for RF (including which metric you use, and how you calculate total decrease due to a feature). In contrast, SHAP is model agnostic when it comes to explaining feature contributions to outcome.
In sum:
Different models will have different opinions about what is important and not. What is important for one algo may be not so important for another and vice versa. It doesn't tell anything about applicability of a model to a specific dataset.
Use SHAP values (or any other feature importance metric that you and your clients understand) to explain a model (if necessary).
Choose "best" model based on your goals: performance or explainability.
Related
For a particular prediction problem, I observed that a certain variable ranks high in the XGBoost feature importance that gets generated (on the basis of Gain) while it ranks quite low in the SHAP output.
How to interpret this? As in, is the variable highly important or not that important for our prediction problem?
Impurity-based importances (such as sklearn and xgboost built-in routines) summarize the overall usage of a feature by the tree nodes. This naturally gives more weight to high cardinality features (more feature values yield more possible splits), while gain may be affected by tree structure (node order matters even though predictions may be same). There may be lots of splits with little effect on the prediction or the other way round (many splits diluting the average importance) - see https://towardsdatascience.com/interpretable-machine-learning-with-xgboost-9ec80d148d27 and https://www.actuaries.digital/2019/06/18/analytics-snippet-feature-importance-and-the-shap-approach-to-machine-learning-models/ for various mismatch examples.
In an oversimplified way:
impurity-base importance explains the feature usage for generalizing on the train set;
permutation importance explains the contribution of a feature to the model accuracy;
SHAP explains how much would changing a feature value affect the prediction (not necessarily correct).
I'm fitting 2 almost identical Random Forest regression models. Both models use the same data set that have 60 features and 90 data points. The only difference is they're using different targets (the target column of each model is excluded from the respective features dataframes, of course). All of the cross validation settings are same of both models (number of folds, number of iterations, scoring) and the hyperparameter grids are also identical.
I'm interested in the feature importance output. However, one of the model consistently output the same top features while the other doesn't. Does anyone know why this is the case?
You can set a seed or the parameter random_state in case you rely on sklearn.ensemble.RandomForestRegressor in order to stabilize your results.
It's quite normal to get varying feature importance since the forest is assembled randomly. Furthermore, feature importance may not be the optimal metric to evaluate actual feature importance. You could try Boruta-Algorithm/Permutation Feature Importance to get a different perspective.
Towards your actual question, maybe your regressors are better suited to predict one target variable over the other.
How do both models perform accuracy-wise on the data? This might be one possibility to explain why one model is more stable. Do feature importances remain unstable for a larger amount of trees fitted?
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 would like to know what are the various techniques and metrics used to evaluate how accurate/good an algorithm is and how to use a given metric to derive a conclusion about a ML model.
one way to do this is to use precision and recall, as defined here in wikipedia.
Another way is to use the accuracy metric as explained here. So, what I would like to know is whether there are other metrics for evaluating an ML model?
I've compiled, a while ago, a list of metrics used to evaluate classification and regression algorithms, under the form of a cheatsheet. Some metrics for classification: precision, recall, sensitivity, specificity, F-measure, Matthews correlation, etc. They are all based on the confusion matrix. Others exist for regression (continuous output variable).
The technique is mostly to run an algorithm on some data to get a model, and then apply that model on new, previously unseen data, and evaluate the metric on that data set, and repeat.
Some techniques (actually resampling techniques from statistics):
Jacknife
Crossvalidation
K-fold validation
bootstrap.
Talking about ML in general is a quite vast field, but I'll try to answer any way. The Wikipedia definition of ML is the following
Machine learning, a branch of artificial intelligence, concerns the construction and study of systems that can learn from data.
In this context learning can be defined parameterization of an algorithm. The parameters of the algorithm are derived using input data with a known output. When the algorithm has "learned" the association between input and output, it can be tested with further input data for which the output is well known.
Let's suppose your problem is to obtain words from speech. Here the input is some kind of audio file containing one word (not necessarily, but I supposed this case to keep it quite simple). You'd record X words N times and then use (for example) N/2 of the repetitions to parameterize your algorithm, disregarding - at the moment - how your algorithm would look like.
Now on the one hand - depending on the algorithm - if you feed your algorithm with one of the remaining repetitions, it may give you some certainty estimate which may be used to characterize the recognition of just one of the repetitions. On the other hand you may use all of the remaining repetitions to test the learned algorithm. For each of the repetitions you pass it to the algorithm and compare the expected output with the actual output. After all you'll have an accuracy value for the learned algorithm calculated as the quotient of correct and total classifications.
Anyway, the actual accuracy will depend on the quality of your learning and test data.
A good start to read on would be Pattern Recognition and Machine Learning by Christopher M Bishop
There are various metrics for evaluating the performance of ML model and there is no rule that there are 20 or 30 metrics only. You can create your own metrics depending on your problem. There are various cases wherein when you are solving real - world problem where you would need to create your own custom metrics.
Coming to the existing ones, it is already listed in the first answer, I would just highlight each metrics merits and demerits to better have an understanding.
Accuracy is the simplest of the metric and it is commonly used. It is the number of points to class 1/ total number of points in your dataset. This is for 2 class problem where some points belong to class 1 and some to belong to class 2. It is not preferred when the dataset is imbalanced because it is biased to balanced one and it is not that much interpretable.
Log loss is a metric that helps to achieve probability scores that gives you better understanding why a specific point is belonging to class 1. The best part of this metric is that it is inbuild in logistic regression which is famous ML technique.
Confusion metric is best used for 2-class classification problem which gives four numbers and the diagonal numbers helps to get an idea of how good is your model.Through this metric there are others such as precision, recall and f1-score which are interpretable.
I'm attempting to make a classifier that chooses a rating (1-5) for a item i. For each item i, I have a vector x containing about 40 different quantities pertaining to i. I also have a gold standard rating for each item. Based on some function of x, I want to train a classifier to give me a rating 1-5 that closely matches the gold standard.
Most of the information I've seen on classifiers deal with just binary decisions, while I have a rating decision. Are there common techniques or code libraries out there to deal with this sort of problem?
I agree with you that ML problems in which the response variable is on an ordinal scale
require special handling--'machine-mode' (i.e., returning a class label) seems insufficient
because the class labels ignore the relationship among the labels ("1st, 2nd, 3rd");
likewise, 'regression-mode' (i.e., treating the ordinal labels as floats, {1, 2, 3}) because
it ignores the metric distance between the response variables (e.g., 3 - 2 != 1).
R has (at least) several packages directed to ordinal regression. One of these is actually called Ordinal, but i haven't used it. I have used the Design Package in R for ordinal regression and i can certainly recommend it. Design contains a complete set of functions for solution, diagnostics, testing, and results presentation of ordinal regression problems via the Ordinal Logistic Model. Both Packages are available from CRAN) A step-by-step solution of an ordinal regression problem using the Design Package is presented on the UCLA Stats Site.
Also, i recently looked at a paper by a group at Yahoo working on ordinal classification using Support Vector Machines. I have not attempted to apply their technique.
Have you tried using Weka? It supports binary, numerical, and nominal attributes out of the box, the latter two of which might work well enough for your purposes.
Furthermore, it looks like one of the classifiers that's available is a meta-classifier called OrdinalClassClassifier.java, which is the result of this research:
Eibe Frank and Mark Hall, A simple approach to ordinal classification. In Proceedings of the 12th European Conference on Machine Learning, 2001, pp. 145-156.
If you don't need a pre-made approach, then these references (in addition to doug's note about the Yahoo SVM paper) might be useful:
W Chu and Z Ghahramani, Gaussian processes for ordinal regression. Journal of Machine Learning Research, 2006.
Wei Chu and S. Sathiya Keerthi, New approaches to support vector ordinal regression. In Proceedings of the 22nd international conference on Machine Learning, 2005, 145-152.
The problems that dough has raised are all valid. Let me add another one. You didn't say how you would like to measure the agreement between the classification and the "gold standard". You have to formulate the answer to that question as soon as possible, as this will have a huge impact on your next step. In my experience, the most problematic part of any (ok, not any, most) optimization task is the score function. Try asking yourself whether all errors equal? Does miss-classifying the "3" as being "4" has the same impact as classifying "4" as "3"? What about "1" vs "5". Can mistakenly missing one case have disastrous consequences (miss HIV diagnosis, activate pilot ejection in a plane)
The simplest way to measure the agreement between categorical classifiers is Cohen's Kappa. More complicated methods are described in the following links here, here, here, and here
Having said that, sometimes picking a solution that "just works", instead of "the right one" is faster and easier. If I were you I would pick a machine learning library (R, Weka, I personally love Orange) and see what I get. Only if you don't have reasonably good results with that, look for more complex solutions
If not interested in fancy statistics a one hidden layer back propagation neural network with 3 or 5 output nodes will probably do the trick if the training data is sufficiently large. Most NN classifiers try to minimize the mean squared error which is not always desired. Support Vector Machines mentioned earlier is a good alternative.
FANN is a good library for back propagation NNs, it also has some tools to assist in training of the network.
There are two packages in R that might help taming ordinal data
ordinalForest on CRAN
rpartScore on CRAN
I'm working on an OrdinalClassifier that is based on the sklearn framework (specifically the OVR multiclass classifier) and which works well with sklearn workflow such as pipelines, cross validation, and scoring.
Through testing, I'm finding that it performs very well vs. standard non-ordinal multiclass classification using SVC. And it gives much greater control over optimizing for precision and recall on the positive class (in my testing, I used sklearn's diabetes dataset and transformed the disease progression target(y) into a low, medium, high class label. Testing via cross validation is on my repo along with attribution. Scoring is based on weighted f1.
https://github.com/leeprevost/OrdinalClassifier