The AiTO is a new and more user friendly way of interacting with people. This is done in two ways,
* Firstly via strategically located help centers to allow more people easier access to the service.
* Secondly with our Speech analysis retrieval assistant or SARA for short, Sara is designed to be an AI call center assistant to not only solve various caller questions while they are in the queue to speak to a representative but to also display the correct information about the caller if it cannot be solved by SARA alone, this is done in an attempt to drastically reduce wait times and processing speed of the ATO call centers.
The following datasets have been used in this project:
* ATO/ABS Data -https://data.gov.au/dataset/govhackato/resource/f3bcbd38-b3e9-4a27-8729-2314f05a6ae4
* GNAF - http://gnafld.net/
* Postcode/LatLong - http://www.corra.com.au/australian-postcode-location-data/
ATO TAX CENTER PREDICTION:
For a complete breakdown oh how the datasets have been used you can view the Notebook file on the github (https://github.com/giskmov/govhack2018_aito/blob/master/govhack_2018.ipynb) which provides a step by step breakdown of how the data has been used and the prediction model constructed.
A brief overview is provided below:
1. Initially the three datasets are cleansed to remove any unnecessary or null values
2. Using primary keys in the datasets these sets are then combined to form a single large table
3. Our label we are trying to predict is the "Count" which is the number of ATO tax help centers per postcode. This label (column) is then removed and placed into a seperate dataset.
4. These two datasets now form our features and labels dataset
5. These datasets are then split and shuffled into training and testing sets for our models
6. A prediction model is then created for use of the data
7. The model is trained on the training set, and validated on the testing set
8. Once the model has reached the required accuracy it is complete and can be used for prediction
9. Based on the data we then predict where the ATO should be placing their help centers
10. A graph is then produced which shows where these centers have been placed
As with all ML Applications the results can be improved, tweaking hyperparamters, modifying underlying data, or including additional data can all be done to help achieve a better result.
AI POWERED OMNI-CHANNEL CHATBOT:
Based on the callers location, we can create a profile based on average income, socio economic index, ethnicity.
On their first interaction with Sarah we will also know their gender to a 98% accuracy.
This information along with historical caller satisfaction ratings is used to create models that help predict:
* Best CSR to send the request to
* Optimise the path through the automated system
* Predict questions and offer
* Measuring customer satisfaction/sentiment
Our omni-bot uses natural language processing powered by machine learning to find insights and relationships in text. The service identifies the language of the text; extracts key phrases, places, people, brands, or events; understands how positive or negative the text is; analyzes text using tokenization and parts of speech; and automatically organizes a collection of text files by topic. Using these APIs, you can analyze text and apply the results in a wide range of applications including voice of customer analysis, intelligent document search, and content personalization for web applications.
This is an embedded Microsoft Office presentation, powered by Office Online.
Evidence of Work
Description of Use: This dataset has been used in conjunction with the other two datasets for prediction of ATO Tax Help Centers.
Description of Use: Used for address locations. This dataset has been used in conjunction with the other two datasets for prediction of ATO Tax Help Centers.
Description of Use: Used to plot the existing locations of ATO help centers to contrast against out predictions for placement optimization. This dataset has been used in conjunction with the other two datasets for prediction of ATO Tax Help Centers.
Check back here once the first checkpoint passes to see the challenges this team has entered.