The Seldon JavaScript API provides the simplest method of integrating Seldon onto web based services. It provides the following methods:

• Events : Called to send to Seldon arbitray event data to be used to build a predictive model
• Prediction : do a real time prediction based on some features

An endpoint to allow the injestion of arbitrary JSON event data.

{% highlight http %} GET /js/event/new {% endhighlight %}

Input

• consumer_key
• json : the url encoded JSON data (if provided this will be assumed to contain the event data)
• jsonpCallback : jsonp callback

If a json parameter is not provied then all parameters will be marshalled into a JSON event other than the reserved parameters: consumer_key, consumer_secret, oauth_token, client and jsonpcallback

If a timestamp field is not provided one will be added.

Example

{% highlight http %} http:///js/event/new?consumer_key=XYZ&user=1&num_rooms=4&postcode=sw1&price=400000&jsonpCallback=j {% endhighlight %}

Assuming the consumer key matches client "client1" this would create an event object similar to below:

{% highlight json %} { "num_rooms":4, "postcode":"sw1", "price":400000, "client":"client1", "timestamp":1421336333669 } {% endhighlight %}

{% highlight http %} GET /js/predict {% endhighlight %}

Input

• consumer_key
• json : the url encoded JSON data
• jsonpCallback : jsonp callback

The JSON input should contain two fields

• meta : optional meta data associated with the prediction request
• data : features for the prediction request

The meta data can at present just contain a provided optional prediction id "puid".

Example

A housing price predictor based on features:

{% highlight json %} { "meta" : { "puid" : 1 }, "data": { "num_bedrooms" : 2, "detached" : true, "postcode" : "SW1" } } {% endhighlight %}

Example

{% highlight http %} http:///js/predict?consumer_key=XYZ&json=%7B%22meta%22%3A%7B%22puid%22%3A%221%22%7D%2C%22data%22%3A%7B%22num_rooms%22%3A2%2C%22detached%22%3Atrue%2C%22postcode%22%3A%22sw1%22%7D%7D&jsonpCallback=j {% endhighlight %}

A response maybe like the following where the "price" field is predicted:

{% highlight json %} { "meta": { "puid": "1", "modelName": "model_prices", "variation": "default" }, "predictions": [ {"prediction":400000,"predictedClass":"1","confidence":1.0} ] } {% endhighlight %}