aimeetup #3: data is oxygen for ml
TRANSCRIPT
Data is oxygen for ML
Cracow, 8th December 2016
Hello, I’m Dima Boyko
Dima BoykoRails developer, Python developer, Data Scientist Software engineer @inFakt
[email protected]/dimaboyko
What computer can do better than human?
What human can do better than computer?
What computer can do better than human?+What human can do better than computer?
”Field of study that gives computers the ability tolearn without being explicitly programmed
1959, Arthur Samuel
What computer can do better thanhuman?+What human can do better thancomputer?
Machine learning
Machine learning is not the future
Machine learning ALVIN
YouTubehttps://youtu.be/ilP4aPDTBPE?t=39
1992 !
Machine learning
BoostML can boost existing products by improving quality and usability of some modules
UnlockUsing ML can unlock newproduct use-cases
Machine learning
Machine learning Drawbacks?
Machine learning Drawbacks?
WOW / WTF ratio
Data Algorithm Insight
Machine learning Usage model
Data
Machine learning Usage model
Data
Using of data
Using of data Red Roof Inn
2 to 3 %of flights were canceled
Using of data Red Roof Inn
500daily
Using of data Red Roof Inn
90 000passengers
Using of data Red Roof Inn
Weather data
Using of data Red Roof Inn
Using of data Red Roof Inn
10%more revenue during season
Using of dataLos Angeles Police Department
Historical
DATA
Analysis & Prediction
Reaction
Using of dataLos Angeles Police Department
Using of dataLos Angeles Police Department
Using of dataLos Angeles Police Department
33% 21%Less thefts Less victims
Using of dataLos Angeles Police Department
Using of dataUPS Cargo Delivery
16,9MDelivered cargos daily
195Countries around the globe
Using of dataUPS Cargo Delivery
Orion
• Mathematical model for operations research• Huge processing power in real time
Using of dataUPS Cargo Delivery
13 000Tons of exhausts less
Using of dataUPS Cargo Delivery
6MLitres less fuelusage during the year
+ Fasterdeliveries
Using of data inFakt Automated Accounting
~50 000Invoices booked monthly by accountants
Using of datainFakt Automated Accounting
AutoAccountingBrief product history
Using of datainFakt Automated Accounting
•
•
•
Data from last year Scikit-learn Infrastructure
AutoAccounting Classification
15% invoices
95% correct
•
•
Data from last year Infrastructure
AutoAccounting Classification
55% invoices
95% correct
AutoAccounting Classification
Keep it simple
3 %Wrong Inconsistent
8/10Human mistake
AutoAccounting
~70%Invoices booked automatically
AutoAccounting
~70%Invoices booked automatically
600 / monthHours saved for creative work
AutoAccounting
Know your DATA!
Auto
What’s next?
Auto
What’s next?
#worldwide #vendor_independent #simple
What’s next?
Open Source ?/OpenAutoX