wazo-cdr-ml-tool
The CDR service allows you to keep track of the activity going on your stack. It's a data source that could be valuable in numerous ways. In this article, we will see how to make value out of your CDR using machine learning.
The CDR dataset
Like all machine learning tasks, we need a so-called dataset which is in our case a dump of the CDR into a CSV file obtained via Portal.
The CDR is a tabular dataset, meaning that we have rows or occurrences, and columns or variables. In the case of the CDR, each row is a call, and each column is an attribute of the calls.
The columns are the following:
- id | integer | Unique identifier
- tenant_uuid | string | Unique identifier for the location
- answered | boolean |
- start | datetime | Date and time at which the call landed on the system
- answer | datetime | Date and time at which the call was answered (empty if called wasn't answered)
- end | datetime | Date and time at which the call hung up
- destination_extension | string | Extension (internal id or phone number) of the destination
- destination_name | string | Name of the destination
- destination_internal_extension | integer | Internal extension of the destination
- destination_internal_context | string | Internal context used
- destination_user_uuid | string | Unique identifier of the reached user
- destination_line_id | integer | User line reached
- duration | integer | Call duration
- call_direction | string | Direction of the call (internal, inbound, outbound)
- requested_name | string | Dialed name
- requested_extension | string | Dialed extension (internal id or phone number) of the requested agent
- requested_context | string | Context corresponding of the requested number
- requested_internal_extension | integer | Internal extension corresponding to the request
- requested_internal_context | string | Internal context corresponding to the request
- source_extension | string | Caller extension (internal id or phone number) of the source agent
- source_name | string | Caller name
- source_internal_extension | string | Caller Internal extension
- source_internal_context | integer | Caller internal context
- source_user_uuid | string | Caller uuid
- source_line_id | integer | Id of the caller line
Create your task
You can create a new task by providing input features and providing output target. See examples in the tasks folder.
Run existing models on it
You can then run pre-made model on this task that have already been optimised to deal with CDR data. The models are in the models folder.
You can run the model of your choice with python make.py {model name} {task name}
You will find a pre-made package in the output folder including:
- A command line tool that takes input from stdin (
main.py) - A flask api to interact with the model
- Hand-made plugins that uses the models to work (available in
package/plugins)