In this talk we will share the idea of developing self guiding application that would provide the most engaging user experience possible using crowd sourced knowledge on a mobile interface. We will discuss and share how historical usage data could be mined using machine learning to identify application usage patterns to generate probable next actions. #h2ony
- Powered by the open source machine learning software H2O.ai. Contributors welcome at: https://github.com/h2oai
- To view videos on H2O open source machine learning software, go to: https://www.youtube.com/user/0xdata
2. In this talk we will share
- Idea of developing self guiding application that would
provide the most engaging user experience possible
using crowd sourced knowledge.
- Discuss and share how historical product usage data
could be mined using machine learning to identify
application usage patterns to generate probable next
actions.
Self Guiding User Experience
3. Why ?
If an app takes more than a few seconds to learn, majority of users are
going to uninstall(Mobile)
Creating that engaging, intuitive initial user experience is challenging,
predominantly constrained by
Complexity of the application
Screen real estate
Domain knowledge, Familiarity
Desktop/Web Experience with steep learning curve looses adoption
4. Why ?
Mine user behavior patterns from crowd sourced application usage data.
Identify High Value Actions/Workflows.
Predict user’s next action based on current/previous actions.
Provide best “Engagement Experience” possible.
Focus on Experience beyond Algorithms and Data
7. The Setting
A mobile photo editing app.
Relatively less complicated – approx. 20 possible actions
Constrained in space – ribbon scroll and searching for actions
The Goal
Create engaging user experience, minimize scrolling and
searching
Predictive Feature Panel and Contextual Window
What ?
8. Crowdsourced Product Usage Data
Each row is a set of actions (like a workflow) performed in an image editing session
Total 100K rows of data, of approx. 20 possible actions
001 002 003
How ?
9. Loose coupling between model creation and consumption
Continuous model development and deployment capability
Create Java POJO for the predictive model
Provide REST API interface to predictive model
Integration into an application
“Once models are deployed to the platform, they can
begin receiving API requests and sending predictions
back to the applications.”
How ?
11. Automated Platform to Build and Scale Smart Data Products
Smart
Data
Product
Smart
Data
Product
Smart
Data
Product
AI – Machine Learning Automation Scalability Visual Intelligence
Smart
Data
Product
11
Dev Framework UX/UI Graphics
Tools, Logs,
Monitoring
Smart Data Product
Store
Smart
Data
Product
Smart
Data
Product
Smart
Data
Product
Smart
Data
Product
Smart
Data
Product
12. REST API – H2O + Steam AI Engine
Training
Dataset
Train Model
Deploy/Scale
Data/Domain Scientist
Smart Apps
Predictive
Model
(Java)
Predictive API
(Jar/WAR file)
Steam
Scoring
Servers
Steam Scoring
ServiceBuilder
Steam
Model
Manager
Dev/Ops Software/Data Engineer
Application Usage Data Collection
13. STEAM – Operationalize Data Science
• Single platform for DevOps, data scientists,
software engineers, and domain scientists to
collaborate on
• Support language of choice for different personas:
R, Python, Java
• Facilitate in-the-moment communication, reduce
model deployment time and get to the results much
faster
• Shared infrastructure with multi-tenancy support
• ElasticML to elastically manage and change the
size of underlying computing cluster
• Reduce your OPEX significantly
Improve Business Efficiency
Improve Operational Resource Efficiency
13
Domain ScientistsData Scientists
Software engineer Data Engineers
DevOps