SQL ANALYTICS – Spark Mllib Algorithm Integration E.G

Oracle Machine Learning

& Advanced Analytics

Data Management Platforms

Move the Algorithms; Not the Data!

Charlie Berger, MS Engineering, MBA

Sr. Director Product Management,

Machine Learning, AI and Cognitive Analytics [email protected]

www.twitter.com/CharlieDataMine

Safe Harbor Statement

The following is intended to outline our general product direction. It is intended for

information purposes only, and may not be incorporated into any contract. It is not a

commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or

functionality described for Oracle’s products remains at the sole discretion of Oracle.

Machine Learning, Data Mining, Predictive Analytics

Automatically sift through large amounts of data to find

hidden patterns, discover new insights and make predictions

A1 A2 A3 A4 A5 A6 A7

• Identify most important factor (Attribute Importance)

• Predict customer behavior (Classification)

• Predict or estimate a value (Regression)

• Find profiles of targeted people or items (Decision Trees)

• Segment a population (Clustering)

• Find fraudulent or “rare events” (Anomaly Detection) • Determine co-occurring items in a “baskets” (Associations)

Oracle’s Data Management and Machine Learning

Market Observations:

• Machine learning, predictive analytics & “AI” now must-have requirements

• Separate islands for data management and data science just don’t work

• Enterprises whose data science teams most rapidly extract insights and

predictions win

Conclusions:

• Must “operationalize” ML insights and predictions throughout enterprise

• Multilingual Machine Learning: SQL, R, Python, Workflow UI, Notebooks, Embed ML in Appls,

• Evolving towards combined Data Management + Machine Learning

environment that can essentially to manage and “think” about data

Oracle’s Data Management and Machine Learning

Architectural Strategy

• In-Database proprietary implementations of machine learning algorithms • Leverage strengths of the Database and adds new ML tech

– Counting, conditional probabilities, sort, rank, partition, group-by, collections, etc. – Parallel execution, bitmap indexes, partitioning, aggregations, recursion w/in parallel infrastructure, IEEE float, frequent itemsets, Automatic Data Preparation (ADP), Text processing, etc.

• Focus on intelligent ML defaults, simplification & automation to enable applications

– ADP, xforms, binning, missing values, Prediction_Details, Predictive_Queries, Model_views

• Machine learning models built via PL/SQL script; scored via SQL functions (1^stclass DB objects)

select cust_id

from customers

True power evident when scoring

where region = ‘US’

and prediction_probability(churnmod, ‘Y’ using *) > 0.8;

models using SQL functions, e.g.

– “Smart scan” ML model scoring “push down”; Supports OLTP and ATPC environments

• Machine Learning and Advanced Analytics are peer to rest of Oracle Data Mgmt features

– Security, Back-up, Encryption, Scalability, “Big Data” eco-system, BDA, Big Data SQL, Cloud, Spark, etc. – Best ML & analytical development and deployment platform

Oracle’s Machine Learning & Advanced Analytics

Fastest Way to Deliver Enterprise-wide Predictive Analytics

Traditional ML
Oracle’s in-DB Machine Learning

Major Benefits

Data Import Data Mining

. Data remains in Database & Hadoop

. Model building and scoring occur in-database

. Use R packages with data-parallel invocations

Model “Scoring”

Data Prep. &
Transformation

avings

. Leverage investment in Oracle IT

. Eliminate data duplication

Data Mining
Model Building

. Eliminate separate analytical servers

Data Prep &
Transformation

. Deliver enterprise-wide applications

. GUI for ML/Predictive Analytics & code gen

. R interface leverages database as HPC engine

Model “Scoring”

Embedded Data Prep Model Building

Data Extraction

Data Preparation

Secs, Mins or Hours
Hours, Days or Weeks

Oracle’s Machine Learning & Adv. Analytics Algorithms

– Non-negative Matrix Factorization – Singular Value Decomposition (SVD) – Explicit Semantic Analysis (ESA)
– Logistic Regression (GLM) – Decision Tree
– Generalized Linear Model – Support Vector Machine (SVM) – Stepwise Linear regression

– Neural Network

– Random Forest

– Neural Network

TEXT MINING SUPPORT

– Support Vector Machine – Explicit Semantic Analysis
– LASSO
– Algorithms support text type – Tokenization and theme extraction – Explicit Semantic Analysis (ESA) for

document similarity

A1 A2 A3 A4 A5 A6 A7

ATTRIBUTE IMPORTANCE
CLUSTERING

– Minimum Description Length

– Principal Comp Analysis (PCA)

– Unsupervised Pair-wise KL Div – CUR decomposition for row & AI

– Hierarchical K-Means

– Hierarchical O-Cluster – Expectation Maximization (EM)

STATISTICAL FUNCTIONS

– Basic statistics: min, max, median, stdev, t-test, F-test, Pearson’s, Chi-Sq, ANOVA, etc.

– State of the art forecasting using
Exponential Smoothing.

through Embedded R Execution
– Spark MLlib algorithm integration

SQL ANALYTICS

– Includes all popular models e.g. Holt-Winters with trends,

seasons, irregularity, missing data

– SQL Windows, SQL Patterns,

SQL Aggregates

EXPORTABLE ML MODELS

• OAA (Oracle Data Mining + Oracle R Enterprise) and ORAAH combined

– REST APIs for deployment

• OAA includes support for Partitioned Models, Transactional, Unstructured, Geo-spatial, Graph data. etc,

Multiple Data Scientist User Roles Supported

Oracle’s Machine Learning/Advanced Analytics

Additional relevant data
Historical or Current Data to

be “scored” for predictions and “engineered features”

Oracle Database 12c

historical data

Sensor data, Text, unstructured data, transactional data, spatial data, etc.

OAA Model Build and Real-time SQL Apply Prediction

Simple SQL Syntax

ML Model Build (PL/SQL)

begin dbms_data_mining.create_model('BUY_INSUR1', 'CLASSIFICATION',
'CUST_INSUR_LTV', 'CUST_ID', 'BUY_INSURANCE', CUST_INSUR_LTV_SET'); end;

Model Apply (SQL query)

Select prediction_probability(BUY_INSUR1, 'Yes'
USING 3500 as bank_funds, 825 as checking_amount, 400 as credit_balance, 22 as age, 'Married' as marital_status, 93 as

MONEY_MONTLY_OVERDRAWN, 1 as house_ownership)

from dual;

Fraud Prediction Demo

Automated In-DB Analytical Methodology

drop table CLAIMS_SET; exec dbms_data_mining.drop_model('CLAIMSMODEL');

create table CLAIMS_SET (setting_name varchar2(30), setting_value varchar2(4000));

insert into CLAIMS_SET values ('ALGO_NAME','ALGO_SUPPORT_VECTOR_MACHINES'); insert into CLAIMS_SET values ('PREP_AUTO','ON'); commit;

begin dbms_data_mining.create_model('CLAIMSMODEL', 'CLASSIFICATION',

'CLAIMS', 'POLICYNUMBER', null, 'CLAIMS_SET');

end; /

Automated Monthly “Application”! Just

-- Top 5 most suspicious fraud policy holder claims select * from (select POLICYNUMBER, round(prob_fraud*100,2) percent_fraud, rank() over (order by prob_fraud desc) rnk from
(select POLICYNUMBER, prediction_probability(CLAIMSMODEL, '0' using *) prob_fraud from CLAIMS

add:

Create

View CLAIMS2_30 As Select * from CLAIMS2 Where mydate > SYSDATE – 30 where PASTNUMBEROFCLAIMS in ('2to4', 'morethan4'))) where rnk <= 5 order by percent_fraud desc;

Time measure: set timing on;

ML Model Deployment for Real-Time Scoring

Real-Time Scoring, Predictions and Recommendations

• On-the-fly, single record apply with new data (e.g. from call center)

Oracle Cloud

Select prediction_probability(CLAS_DT_1_15, 'Yes'

USING 7800 as bank_funds, 125 as checking_amount, 20 as credit_balance, 55 as age, 'Married' as marital_status, 250 as MONEY_MONTLY_OVERDRAWN, 1 as house_ownership)

from dual;

Social Media

Likelihood to respond:

Call Center
Branch

Office

Get Advice

Mobile
Web

Manage and Analyze All Your Data

Data Scientists, R Users, Citizen Data Scientists

Architecturally,

Many Options

and Flexibility

SQL / R

Boil down the Data Lake

“Engineered Features”

Big Data SQL / R

– Derived attributes that reflect domain

knowledge—key to

best models e.g.:
• Counts

Object

Store

• Totals • Changes

over time

OAA Algorithm Performance

SVM SGD Solver GLM Classification SVM IPM Solver

GLM Regression

K-Means
83s
180s 811s

900
9000
900

• In-memory 640 million rows, Airline On-Time dataset • SPARC M8-2 hardware • Courtesy of the SPARC Performance Team

OAA Algorithm Scalability: Rows and Parallelism

Random Forest

200 180

140 120 100

60 40 20
0

Row Count (M records)

• X6-2 hardware • Benchmarks from Oracle Labs, Systems
Research Group

Operationalizing and Embedding Analytics for Action

How long does it take to put a defined model into operational use?

Length of time to put a model into production.

Based on 141 respondents who

stated they are doing this today

The Core Ingredients of Good Machine Learning

Machine Learning Algorithms

A1 A2 A3 A4 A5

A1 A2 A3 A4 A5 A6 A7

Deployment!!

Most Important Factor in Machine Learning?

Machine Learning Algorithms

A “Thinking”

Database

A1 A2 A3 A4 A5 A6 A7

Introducing:

Oracle Machine Learning SQL Notebook

Introducing Oracle Autonomous Data Warehouse Cloud

Value Proposition

• Provision a data warehouse in as little

defined sizing, on-demand scaling & idle shut-off

as 15-seconds
• Automated management of database

Redshift¹
• High concurrency supports multi-user

administration

• Simple Load and Go with Automated

access and workloads

• Based on Exadata for extreme

• Independent scaling of compute and

storage

Tuning

• Dedicated cloud-ready migration tools including Redshift

Confidential – Oracle

Oracle Autonomous Data Warehouse Cloud Key Features

Highly Elastic
High-Performance Queries

Independently scale compute and storage, without having to overpay for fixed blocks of resources

and Concurrent Workloads

Optimized query performance with

preconfigured resource profiles for different types of users

Oracle Machine Learning

Oracle SQL

Built-in Web-Based SQL ML Tool

Apache Zeppelin Oracle Machine Learning

notebooks ready to run ML from browser
Autonomous DW Cloud is compatible with all business analytics tools that support

Oracle Database

Database migration utility
Self Driving

Dedicated cloud-ready migration tools

for easy migration from Amazon
Fully automated database for self-tuning patching and upgrading itself while the system is running
Redshift, SQL Server and other databases

Cloud-Based Data Loading

Fast, scalable data-loading from Oracle Object Store, AWS S3, or on-premises

Enterprise Grade Security

Data is encrypted by default in the cloud, as well as in transit and at rest

Confidential – Oracle

ADWC + Oracle Machine Learning Notebooks

Developer
Analytics

Autonomous Data Warehouse Cloud

Tools

Data Warehouse Services

(EDWs, DW, departmental marts and sandboxes)

Oracle Exadata Cloud Service

Oracle Analytics Cloud
Oracle SQL Developer

Oracle Database Cloud Service

Service Console

Oracle Machine Learning

Express Cloud Service

Data
Integration
Services

Oracle Data Integration
Platform Cloud

Autonomous Database Cloud

3^rdParty DI on
Oracle Cloud Compute

Oracle Object Storage Cloud

Flat Files and Staging

3^rdParty DI On-premises

Machine Learning Notebook for Autonomous Data Warehouse Cloud

Key Features

• Collaborative UI for data scientists

– Packaged with Autonomous Data

Warehouse Cloud (V1)
– Easy access to shared notebooks, templates, permissions, scheduler, etc.

– SQL ML algorithms API (V1)

– Supports deployment of ML analytics

Machine Learning Notebook for Autonomous Data Warehouse Cloud

Key Features

• Collaborative UI for data scientists

– Packaged with Autonomous Data

Warehouse Cloud (V1)
– Easy access to shared notebooks, templates, permissions, scheduler, etc.

– SQL ML algorithms API (V1)

– Supports deployment of ML analytics

Oracle’s Machine Learning & Adv. Analytics Algorithms

– Non-negative Matrix Factorization – Singular Value Decomposition (SVD) – Explicit Semantic Analysis (ESA)
– Logistic Regression (GLM) – Decision Tree
– Generalized Linear Model – Support Vector Machine (SVM) – Stepwise Linear regression

– Neural Network

– Random Forest

– Neural Network

STATISTICAL FUNCTIONS

– Basic statistics: min, max, median, stdev, t-test, F-test, Pearson’s, Chi-Sq, ANOVA, etc.
– Support Vector Machine – Explicit Semantic Analysis

A1 A2 A3 A4 A5 A6 A7

ATTRIBUTE IMPORTANCE

– Minimum Description Length

– Principal Comp Analysis (PCA)

– Unsupervised Pair-wise KL Div

CLUSTERING

– Hierarchical K-Means

– Hierarchical O-Cluster – Expectation Maximization (EM)

ASSOCIATION RULES
ANOMALY DETECTION

– A priori/ market basket
– One-Class SVM

SQL ANALYTICS – Spark Mllib Algorithm Integration E.G

Details

Download

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

Support