6.14 Takeaways

This chapter explains fusions in detail. Here is a list of key points:

• A fusion is a supervised Machine Learning algorithm used to solve classification and regression problems.

• A fusion is built by selecting multiple models in BigML and using to make an evaluation, a single prediction, or a batch prediction (see Figure 6.87 ).

• You can create a fusion using models, ensembles, logistic regression and/or deepnets.

• You can assign different weights to the models composing a fusion.

• All models need to have the same Objective Field.

• The PDP view provides a visual way to analyze a field’s impact on predictions given certain values for the rest of the fields.

• For classification problems, you get all the objective field class probabilities along with the predicted class.

• For regression problems, you get the objective field predicted values.

• BigML lists all the models composing the fusion along with their weights.

• You need to evaluate your fusion’s performance using data that the fusion has not seen before.

• The ultimate goal in building a fusion is being able to make predictions with it.

• BigML allows you to quickly make predictions for single instances by providing a form containing the fields used by the fusion, so you can easily set the input data and get an immediate response.

• BigML batch predictions allow you to make simultaneous predictions for multiple instances. All you need is the fusion you want to use to make predictions and a dataset containing the instances for which you want to obtain predictions.

• You can configure your batch predictions output file settings.

• You can download your fusion to perform local predictions.

• You can add descriptive information to your fusion (name, description, tags, and category).

• You can move your fusions between projects.

• You can share your fusions with other people using the secret link.

• You can stop your fusions creation by deleting them.

• You can permanently delete an existing fusion.