Anomaly Detection with the BigML Dashboard

6.4 Visualizing Anomaly Scores

The visualization of anomaly scores is different for single anomaly scores and batch anomaly scores. The following sections explain both of them.

6.4.1 Single Anomaly Scores

When scoring single instances, you can find the anomaly score at the top of the form. (See Figure 6.23 .) Remember that instances scoring higher than 60% may be considered anomalous. You can change the value of the input fields at any time and click Score to have your score recalculated.

\includegraphics[]{images/an-visualize-single}
Figure 6.23 Single scores view

All your scores will be saved and listed in the score list view. (See Figure 6.1 .)

6.4.2 Batch Anomaly Scores

For batch scores, you always get an output file and, optionally, an output dataset.

Output File

From the batch score view you can access the file view as shown in Figure 6.24 .

\includegraphics[]{images/batchpred-anomalies-csv}
Figure 6.24 Download batch prediction output file

By default, it will be a CSV file containing all the dataset fields. The last column includes the predicted scores for each of the instances. You can customize the output file settings as explained in subsection 6.3.2 .

See an output CSV file example in Figure 6.25 where the last column contains the anomaly score for each instance.

Pregnancies,Glucose,Insulin,BMI,Diabetes pedigree,Age,Diabetes,score
0,137,168,43.1,2.288,33,TRUE,0.5332
11,143,146,36.6,0.254,51,TRUE,0.49067
4,103,192,24.0,0.966,33,FALSE,0.46055
7,106,0,22.7,0.235,48,FALSE,0.42654
2,109,0,42.7,0.845,54,FALSE,0.47717
1,122,220,49.7,0.325,31,TRUE,0.51672
8,155,495,34.0,0.543,46,TRUE,0.51943
4,76,0,34.0,0.391,25,FALSE,0.38641
1,118,94,33.3,0.261,23,FALSE,0.36354
Figure 6.25 An example of a batch prediction CSV file

Output Dataset

By default BigML automatically creates a dataset out of your batch anomaly score. You can disable this option by configuring your batch score as explained in section 6.3 . You can access your output dataset from the batch score view as shown in Figure 6.26 .

\includegraphics[]{images/batchpred-anomalies-dataset}
Figure 6.26 Access batch prediction output dataset

In the output dataset you can find an additional field (named by default “score”) containing the anomaly score for each one of your instances. (See Figure 6.27 .) If you configured your batch score to include the field importance, you will be able to find an additional field for each field in your anomaly detector appended at the end of your output dataset.

\includegraphics[]{images/dataset-out}
Figure 6.27 Batch scores output dataset

Batch Scores 1-Click Action Menu

From the batch score view you can perform the following actions shown in Figure 6.28

  • BATCH SCORE AGAIN: this option will redirect you to the batch score creation view where you will have the same anomaly and score dataset already selected. It is a quick way if you want to create the batch score again using a different configuration.

  • BATCH SCORE WITH ANOTHER DATASET: this is an easy way to create a batch score using the same anomaly and a different dataset.

  • BATCH SCORE USING ANOTHER ANOMALY: this is an easy way to create a batch score using the same dataset and a different anomaly.

  • NEW BATCH SCORE: this will redirect you to the batch score creation view where you will be able to select a score dataset and a anomaly to create your batch score.

\includegraphics[]{images/1-click-actions}
Figure 6.28 Batch score 1-click action menu