1mlsac.jpg

Source: SAP

This was an SAP webcast from last week.  Tomorrow there is a related webcast called Smart Predict - register here

2mlsac.jpg

Source: SAP

"We have to do more with less"

3mlsac.jpg

Source: SAP

How embed machine learning in the application

How get the information to the right decision at the right time in the right context

4mlsac.jpg

Source: SAP

Data discovery  has failed to meet expectations

Analytics is subject to human bias; selection bias as you can’t ingest all information

How get to insights; typically through visual recognition; takes too much time causing delayed insights

 

Next step to solve; plan, use machine learning – eliminates bias, uncovers insights; machine learning recommends next action to take

 

Adoption of machine learning in analytics is quite low; usually machine learning is geared towards data scientists

 

How make machine learning accessible to the BI user

5mlsac.jpg

Source: SAP

Planning is fragmented, ignored, reactive

 

Organizations do planning with spreadsheets today, end up with version control issues, and planning process breaks down, and takes 3-6 months and gets ignored

 

Bring planning with machine learning and analytics, you can monitor and plan

6mlsac.jpg

Source: SAP

Not every user will adopt machine learning models

 

Make it easier to get answers naturally; use natural language

 

Explain a measure with a single click; what drives revenue, use machine learning in background

 

Human does not need to code a model

 

If a business analyst knows it is a classification model (will customer buy x) they can do it and will know if model is quality.

7mlsac.jpg

 

Source: SAP

Going from data to insights as quickly; address data quality

8mlsac.jpg

Source: SAP

Type in questions, like "what is the revenue for 2019" and then see the results in a visualization

Users can see visualizations without selecting chart type, measure

Increase abilities of business analyst

9mlsac.jpg

Source: SAP

Help automatically interesting patterns in data, influencers, drivers of revenue, determine anomalies in dataset, where are the outliers

 

Simulate drivers, what are the impacts

10mlsac.jpg

Source: SAP

 

Detect where you might have challenges - an early warning system

Click of a button with time series analysis, you can see if risk to make target

11mlsac.jpg

Source: SAP

Classification is Y/N – will someone buy?  Churn or not?

Regression – how explain something? What are variables contributing to measure

Time series – here what has happened historically and project to future, and if trend continues, where end up

 

Smart features with pre-built content (15 industries) and 27 line of business use cases (finance, HR, marketing)

12mlsac.jpg

 

 

Source: SAP

Plan head count expenses, where allocate

Work with business

Collaborate with stakeholders

Come up with plan

Trends on expenses

Enhance where employees may churn

Track performance against key KPI's and embed insights into HR SuccessFactors

 

Question & Answer

Q: What does predictive power and confidence mean?

A: An indication of 2 qualities

Predictive power of a model - how much of the thing you are trying to explain, were we able to explain based on data provided

Confidence - how apply to new data and be valid

 

Prediction confidence is more critical

 

Q: Difference from machine learning?

Machine learning foundation can be used by developers