It became clear quickly that it is desirable to cut down on the number of taste and odour related calls. This can be done by reducing the need for someone to call but cannot be done by making it harder to call.

Our idea was to identify an issue affecting multiple users in real-time and inform the customer of this as part of the contact process so they will have a reduced need to call.

Our solution uses the data from incoming calls to identify the larger issues affecting others in real-time. This allows the web site to be automatically updated so a customer knows their issue is already known about and being addressed. This will only happen when a number of customers have already called so no important information will be lost.

This solution processes contacts made from the general public as they are entered on to the Oracle system. Each contact is linked together with other information including the geolocation, DMA and pipes likely to be used to deliver the water. A pattern search is then done on this data to look for other recent issue with a common element.

When a potential issue is identified a notification is shown and the location plotted on the map. When a common element to an issue is identified the web site contact page can be automatically updated to inform the customer before they call.

The prediction engine is written as a command line application in PHP. For the demonstration we are using a PHP script to run thought the past call to simulate a real file scenario. The base engine use MySQL as a cache for the data. Then an issue is identified Pusher is used to notify the other element for the project. The front end display is written in JQuery and receives the updates to show on a Leaflet driven map.

The current prototype is slightly limited by the base data se can link in to it because of security considerations. Ideally we would be able to trial the solution with un-redacted data to allow it to be tested fully.

We would also like to link in more historical and live information to allow the likelihood and severity of an issue to be better identified.

Finally being able to integrate live web analytics will further help identify and quantify issues.

Share this project: