The PCA model is used to find the linear combination of your original features that best describes your data. In that sense, the goal of the model is to provide a transformation that allows dimensionality reduction. Following the schema described in the prediction workflow document, this is the code snippet that shows the minimal workflow to create a PCA model and produce a single projection.
;; step 0: creating a source from the data in your remote "https://static.bigml.com/csv/iris.csv" file
(define source-id (create-source {"remote" "https://static.bigml.com/csv/iris.csv"}))
(log-info "Creating remote source: " source-id)
;; step 1: creating a dataset from the previously created `source`
(define dataset-id (create-dataset source-id))
(log-info "Creating dataset from source: " dataset-id)
;; step 2: creating a decision tree model
(define pca-id (create-pca dataset-id))
(log-info "Creating PCA from dataset: " pca-id)
;; the new input data to predict for
(define input-data {"petal width" 1.75
"petal length" 2.45
"sepal length" 4
"sepal width" 1})
;; creating a single projection
(define projection-id (create-projection pca-id
{"input_data" input-data}))
(log-info "Projection has been created: " projection-id)
(define projection-resource (fetch projection-id))
;; extracting the projection components from the projection-resource
(define projection (projection-resource ["projection" "result"]))
(log-info "Projection: " projection)
You can test this code in the WhizzML REPL.
Note that create calls are not waiting for the asynchronous resources to be
finished before returning control. However, when a resource is used to generate
another one, the next create call will wait for the origin resource to be
finished before starting the creation process (e.g, (create-dataset source-id)
will wait for the source to be finished before creating the dataset from it).
If you want to create projections for all the rows in the dataset, you can do
so by creating a batchprojection resource.
;; step 3: creating a batch projection
(define batch-projection-id (create-batchprojection pca-id
dataset-id))
The batch projection output (as well as any of the resources created)
can be configured using additional arguments in the corresponding create calls.
For instance, to include all the information in the original dataset in the
output you would change step 3 to:
(define batch-projection-id (create-batchprojection
{"pca" pca-id
"dataset" dataset-id
"all_fields" true}))
and to generate a new dataset that adds the projection components as new columns appended at the end of the original ones and retrieve the information therein, the step would be:
;; step 3: creating a batch projection. Note that we now use
;; `create-and-wait-batchprojection` to ensure that we wait for the
;; resource to be finished to read its properties in the next step.
;; We can also use `create-batchprojection` and wait later for the
;; resource to finish using `wait`. The `create-[resource-type]` and
;; `create-and-wait-[resource-type]` procedures are available for
;; all resources
(define batch-projection-id (create-and-wait-batchprojection
{"pca" pca-id
"dataset" dataset-id
"all_fields" true
"output_dataset" true}))
;; step 6: retrieving the batch projection information
(define batch-projection-resource (fetch batch-projection-id))
;; step 7: extracting the output dataset ID and waiting for it to be
;; finished too in order to use it.
(define batch-projection-ds
(wait (batch-projection-resource "output_dataset_resource")))
Check the API documentation to learn about the available configuration options for any BigML resource.