I have a classifier for a NER task, and since 'O' labels are by far more than all others, I want to exclude it in metrics calculation.
I want to compute macro and micro scores with sklearn package. Macro scores can be calculated with precision_recall_fscore_support, because it returns the precision, recall, F1 and support for each label separetly.
Can I use sklearn package to compute and micro scores as well?
The answer turns out to be very simple. The label parameter of the function determines which labels to include in scores calculation. It is also combined with the macro, micro averages.
Related
im try to study confusion matrix. i know about 2x2 confusion matrix but i still don't understand how to count 5x5 confusion matrix for finding accuracy, precision, recall and, f1 - score. Can anyone help me with this ? i appreciate every help.
See my answer here: Calculating Equal error rate(EER) for a multi class classification problem
In short, one strategy is to split the multiclass problem into a set of binary classification, for each class a "one vs. all others" classification. Then for each binary problem you can calculate F1, precision and recall, and if you want you can average (uniformly or weighted) the scores of each class to get one F1 score which will represent the multiclass problem.
As for confusion matrix larger than 2x2: the rows are the true labels and the columns are predicated labels. Then the number in cell (i,j) is the number of samples from class i which were classified as class j (note that i=j corresponds to correct prediction). The accuracy is the trace of the confusion matrix divided by the number of samples.
I am trying to implement a scoring model following the link https://rstudio-pubs-static.s3.amazonaws.com/376828_032c59adbc984b0ab892ce0026370352.html#1_introduction.
Post the entire implementation though, When I create pivot with my generated scores and the original labels, the average scores for "good' labels is significantly lower than the ones for " high" labels.
Hence, my problem can be oversimplified to why would logistic regression give reversed probabilities for 0-1 target variable( In my model I am using 0 for bad and 1 for good).
Any suggestions and solutions would be welcome.
I have constructed a GMM-UBM model for the speaker recognition purpose. The output of models adapted for each speaker some scores calculated by log likelihood ratio. Now I want to convert these likelihood scores to equivalent number between 0 and 100. Can anybody guide me please?
There is no straightforward formula. You can do simple things like
prob = exp(logratio_score)
but those might not reflect the true distribution of your data. The computed probability percentage of your samples will not be uniformly distributed.
Ideally you need to take a large dataset and collect statistics on what acceptance/rejection rate do you have for what score. Then once you build a histogram you can normalize the score difference by that spectrogram to make sure that 30% of your subjects are accepted if you see the certain score difference. That normalization will allow you to create uniformly distributed probability percentages. See for example How to calculate the confidence intervals for likelihood ratios from a 2x2 table in the presence of cells with zeroes
This problem is rarely solved in speaker identification systems because confidence intervals is not what you want actually want to display. You need a simple accept/reject decision and for that you need to know the amount of false rejects and accept rate. So it is enough to find just a threshold, not build the whole distribution.
I am using Weka IBk for text classificaiton. Each document basically is a short sentence. The training dataset contains 15,000 documents. While testing, I can see that k=1 gives the best accuracy? How can this be explained?
If you are querying your learner with the same dataset you have trained on with k=1, the output values should be perfect barring you have data with the same parameters that have different outcome values. Do some reading on overfitting as it applies to KNN learners.
In the case where you are querying with the same dataset as you trained with, the query will come in for each learner with some given parameter values. Because that point exists in the learner from the dataset you trained with, the learner will match that training point as closest to the parameter values and therefore output whatever Y value existed for that training point, which in this case is the same as the point you queried with.
The possibilities are:
The data training with data tests are the same data
Data tests have high similarity with the training data
The boundaries between classes are very clear
The optimal value for K is depends on the data. In general, the value of k may reduce the effect of noise on the classification, but makes the boundaries between each classification becomes more blurred.
If your result variable contains values of 0 or 1 - then make sure you are using as.factor, otherwise it might be interpreting the data as continuous.
Accuracy is generally calculated for the points that are not in training dataset that is unseen data points because if you calculate the accuracy for unseen values (values not in training dataset), you can claim that my model's accuracy is the accuracy that is been calculated for the unseen values.
If you calculate accuracy for training dataset, KNN with k=1, you get 100% as the values are already seen by the model and a rough decision boundary is formed for k=1. When you calculate the accuracy for the unseen data it performs really bad that is the training error would be very low but the actual error would be very high. So it would be better if you choose an optimal k. To choose an optimal k you should be plotting a graph between error and k value for the unseen data that is the test data, now you should choose the value of the where the error is lowest.
To answer your question now,
1) you might have taken the entire dataset as train data set and would have chosen a subpart of the dataset as the test dataset.
(or)
2) you might have taken accuracy for the training dataset.
If these two are not the cases than please check the accuracy values for higher k, you will get even better accuracy for k>1 for the unseen data or the test data.
There is a module named SMOTE(Synthetic Minority Oversampling Technique ) which increase the number of samples of under sampled data, I guess we should choose a feature(feature to be predicted) which is under represented. How to choose it? There seems to be no option on choosing the coloumn.
I guess you are referring to the target variable (label column). You can set that using a Metadata Editor module. Choose your label column using the column selector and set the Fields property to Labels.
Here is the SMOTE definition - SMOTE is an approach for the construction of classifiers from imbalanced datasets, which is when classification categories are not approximately equally represented. The classification category is the feature that the classifier is trying to learn. There is not an option of choosing the column in the SMOTE module because it should be the label column
Here is the details on how to use SMOTE in Azure Machine Learning - https://msdn.microsoft.com/en-us/library/azure/dn913076.aspx?f=255&MSPPError=-2147217396
You can do it thru the column selector. In the sample below, the blood donation data (a sample dataset in Azure ML) has 25% of people who donated (class 1).