What is a prediction engine?

1. real predictive power comes from simplicity

1.1. automate stuff in the engine

1.2. flexible queries

1.3. accuracy & honesty

1.4. provide actionability

2. case study 2: optimizing sales process with predictive

2.1. pulled 23 data marts into a "sales data master"

2.2. then optimize the Sales Master process

2.2.1. created a feedback loop

2.2.1.1. measure

2.2.1.2. act

2.2.1.3. compliment data

2.2.1.4. measure again

2.3. with p(k) engine

2.3.1. predict simultaneously

2.3.2. discover

2.3.3. suggest

2.3.4. adapt

3. case-study 1: US health care

3.1. measured many features in US health care

3.2. fed into the engine, to find correlated features, what can predict quality of care

3.2.1. found related output scores

3.2.2. but no correlation to input variables

3.2.3. that'r problematic - means we're not measuring the right things

3.2.4. see visualization slide (colors table)

3.3. P(K) engine streamline the process

3.3.1. Ingest data (MD driven)

3.3.2. Build statistical index (automatic)

3.3.3. Query flexible (DB inspired)

4. cheat-sheet of scikit-learn

4.1. see in slides

5. Today predictive analytics are problematic

5.1. take lots of time

5.2. complicated

5.2.1. many steps in the process & need to start over if any step breaks

5.3. brittle

5.3.1. break as soon as something changes

6. Not just predictions

6.1. Predict

6.2. Suggest

6.3. Discover

6.4. Adapt

7. Dr. Beau ...,

7.1. founded

7.1.1. Prior Knowledge

7.1.1.1. p(k)

7.1.2. acquired by SFDC