This document discusses predictive APIs and machine learning. It covers the two phases of machine learning: training a model and predicting with a model. It also discusses the different types of predictive APIs based on their level of abstraction, including generic, text classification, problem-specific, and fixed model APIs. It provides examples of APIs for each type. The document also discusses customizing predictive APIs and introduces PredictionIO as an open source predictive serving engine.

1. @louisdorard
#apidays

5. PredictiveAPIs

6. –Mike Gualtieri, Principal Analyst at Forrester
“Predictive apps are
the next big thing
in app development.”

7. –Waqar Hasan, VISA
“Predictive is the ‘killer app’ for
big data.”

8. 1. Machine Learning
2. Data

9. BUT

10. –McKinsey & Co. (2011)
“A significant constraint on
realizing value from big data
will be a shortage of talent,
particularly of people with
deep expertise in statistics and
machine learning.”

13. What the @#?~%
is ML?

15. “Which type of email is this?
— Spam/Ham”
-> Classification

16. “Which type of email is this?
— Spam/Ham”
I
O

18. • Need examples of inputs AND outputs
• Need enough examples

19. ??

20. Predictive APIs

21. The two phases of machine learning:
• TRAIN a model
• PREDICT with a model

22. The two methods of predictive APIs:
• TRAIN a model
• PREDICT with a model

23. The two methods of predictive APIs:
• model = create_model(dataset)
• predicted_output =
create_prediction(model, new_input)

24. from bigml.api import BigML
# create a model
api = BigML()
source =
api.create_source('training_data.csv')
dataset = api.create_dataset(source)
model = api.create_model(dataset)
# make a prediction
prediction = api.create_prediction(model,
new_input)
print "Predicted output value:
",prediction['object']['output']
http://bit.ly/bigml_wakari

25. from bigml.api import BigML
# create a model
api = BigML()
source =
api.create_source('training_data.csv')
dataset = api.create_dataset(source)
model = api.create_model(dataset)
# make a prediction
prediction = api.create_prediction(model,
new_input)
print "Predicted output value:
",prediction['object']['output']
http://bit.ly/bigml_wakari

27. Different needs,
different Predictive APIs

28. ABSTRACTION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

29. ABSTRACTION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

30. “Is this email important?
— Yes/No”

31. Generic Predictive APIs:
• Google Prediction API
• BigML
• PredicSis
• WolframCloud
• Amazon

32. AMAZON GOOGLE PREDICSIS BIGML
ACCURACY 0.862 0.743 0.858 0.790
TRAINING
TIME
135s 76s 17s 5s
TEST TIME 188s 369s 5s 1s
louisdorard.com/blog/machine-learning-apis-comparison

33. ABSTRACTION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

34. “Which project does this document relate to?
— A/B/C”

35. Text classification APIs:
• uClassify.com
• MonkeyLearn.com
• Cortical.io

36. ABSTRACTION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

37. “Is this customer going to leave next month?
— Yes/No”

38. “Is this transaction fraudulent?
— Yes/No”

39. Problem-specific Predictive APIs:
• Churn: ChurnSpotter.io, Framed.io
• Fraud detection: SiftScience.com
• Lead scoring: Infer.com
• Personal assistant (Siri): Wit.ai

40. ABSTRACTION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

41. “What is the sentiment of this tweet?
— Positive/Neutral/Negative”

42. “Is this email spam?
— Yes/No”

43. Fixed-model Predictive APIs:
• Text: Datumbox.com, Semantria.com
• Vision: Indico, AlchemyAPI.com
• Siri-like: maluuba.com

44. ABSTRACTION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

45. CUSTOMIZATION
Algorithmic
Generic
Generic for text
Problem-specific
Fixed-model

46. Own ML
algorithm?

47. Own data processing
algorithm?

48. Data processing
APIfication

52. MLlib

55. • “Open source prediction server”
• Based on Spark
• Expose predictive models as (scalable
& robust) APIs

56. = PAPI+

57. • Send (new) data
• Send prediction queries

58. cli = predictionio.EventClient("<my_app_id>")
cli.record_user_action_on_item("view","John","Item")
eng = predictionio.EngineClient("<my_engine_url>")
rec = eng.send_query({"uid":"John","n":5})

60. Use cases

61. • Increase e-commerce revenue with recos
• Reduce attrition (“churn”)
• Automate decision making
• Predict real-estate prices w. open data
• Predict crowds (SNCF & La Poste)
• …

63. papis.io/connect
(discount code: APIDAYS)