The basic process to create a model in BigML goes through several steps. In each step, a new resource is generated. You can see a sketch of the process in the following image.
The first step is creating a Source by uploading your data.
This means that you'll need to use the create-source call
to create a Source object. The only mandatory argument
is the path to your data, and the rest of arguments that can be found in the
API Documentation are
optional. As WhizzML is run in the platform, the path to your data
is usually a URL and it should be provided using the remote argument.
(create-source {"remote" "https://static.bigml.com/csv/iris.csv"})
The result of the create-source call is the new Source ID. All the resources in BigML can be identified using a unique resource ID and the format of this ID for Sources would be source/ followed by an alphanumeric string (e.g.: source/4f603fe203ce89bb2d000000).
This will be achieved by using the create-dataset call and the only mandatory argument for this call is the Source ID which points to the Source object we want to create the Dataset from. The rest of arguments described in the API Documentation are optional.
;; defining a source-id variable that stores the Source ID
(define source-id "source/4f603fe203ce89bb2d000000")
;; creating the dataset from that source
(create-dataset source-id)
The result of the create-dataset call is the corresponding resource ID for the dataset (e.g. dataset/4f603fe203ce89bb2d000004). The dataset summarizes and serializes the entire set of values that each field contains.
The model (cluster, anomaly, etc.) can be created using the configuration options by default or customizing some values to improve its performance. As an example, we'll talk about how to create a decision tree and for that you'll need to use the create-model call. The only madatory argument is the dataset ID to start from (in supervised models, the objective field will be set to the dataset's last field if not explicitly set). The rest of creation arguments described in the API Documentation are optional. For instance, when building a classification model you might like to limit the number of nodes in your decision tree. Also, if the number of instances that belong to each class in your data is very skewed, you should consider balancing them. This can be achieved using a balance_objective argument, as explained in the API Documentation. Also, a major attribute to be taken into account when doing classification or regression is the objective_field, which chooses the field that will be predicted. Any of these properties cannot be updated. If you want to change them, you'll need to create a new model with the new configuration.
;; defining a dataset-id variable that stores the Dataset ID
(define dataset-id "dataset/4f603fe203ce89bb2d000004")
;; creating a decision tree from that dataset
(create-model dataset-id)
The result of a create-model call is a model ID (e.g. model/4f603fe203ce89bb2d0000ac).
Any other modeling resource (like clusters, anomaly detectors, etc.) will also have its particular creation calls (described in their corresponding 101 documents) and configuration arguments (described in their corresponding API documentation section). Again, to use a different configuration you'll need to create a new resource with it.
When your model is ready and you want to create predictions for a new bunch of test data, you will need to run through it your test data file. The process is sketched in the following image:
As you see, steps 3 and 4 (the equivalent of steps 0 and 1 in the training
workflow) are needed too to create the test dataset,
and in step 5 you create a batch prediction
(or batch centroid, batch anomaly score, etc. depending on the model)
with the results of running the test data through the model and adding its
output to it.
Following the decision tree example, batch predictions can be created using the create-batchprediction call. The mandatory arguments for the create-batchprediction call are the ID of the model used to predict and the ID of the dataset that contains the test instances that you want to predict for. The rest of arguments described in the API Documentation are optional.
;; defining a test-dataset-id variable that stores a Dataset ID to be used as input data
(define test-dataset-id "dataset/4f603fe203ce89bb2d000009")
;; defining a model variable that stores the decision tree ID
(define model-id "model/4f603fe203ce89bb2d0000ac")
;; creating a batch prediction by running every instance in the test dataset
;; through the decision tree
(create-batchprediction {"model" model-id
"dataset" test-dataset-id})
The result of a create-batchprediction call is a batchprediction ID (e.g. batchprediction/4f603fe203ce89bb2d0000fd).