History and Trend of Big Data and Deep Learning

Jongwook Woo
HiPIC
CalStateLA
Keimyung University
Dec 20 2019
Jongwook Woo, PhD, jwoo5@calstatela.edu
Big Data AI Center (BigDAI)
California State University Los Angeles
History and Trend of
Big Data and Deep Learning

Big Data Artificial Intelligence Center (BigDAI)
Jongwook Woo
CalStateLA
Contents
 Myself
 Introduction To Big Data
 Deep Learning and Big Data
 Big Data Predictive Analysis
 Summary

Jongwook Woo
CalStateLA
Myself
Experience:
Since 2002, Professor at California State University Los Angeles
– PhD in 2001: Computer Science and Engineering at USC

Jongwook Woo
CalStateLA
Universities in Los Angeles
West
North

Jongwook Woo
CalStateLA
Universities in Los Angeles

Jongwook Woo
CalStateLA
California State University
Los Angeles

Jongwook Woo
CalStateLA
Myself: S/W Development Lead
http://www.mobygames.com/game/windows/matrix-online/credits

Jongwook Woo
CalStateLA
Collaboration with HDP, CDH, Oracle, Amazon
using Hadoop Big Data
https://www.cloudera.com/more/customers/csula.html

Jongwook Woo
CalStateLA
Myself: Partners for Services

Jongwook Woo
CalStateLA
Myself: Collaborations

Jongwook Woo
CalStateLA
New Technology: Big Data
What is Big Data? Data or Systems?
Large Scale Data?
–Many people only see the data point of view
–3 Vs, 5Vs
Systems?
– YES

Jongwook Woo
CalStateLA
Data Handling Systems: Traditional Way

Jongwook Woo
CalStateLA
Data Handling: Traditional Way
Becomes too Expensive

Jongwook Woo
CalStateLA
Data Handling: Another Way
Not Expensive
From 2017 Korean
Blockbuster Movie,
“The Fortress”
(남한산성)

Jongwook Woo
CalStateLA
Data Handling: Another Way
Not Expensive
http://blog.naver.com/PostView.nhn?blogId=dosims&logNo=221127053677
1409년(태종 9) 최해산(崔海山), 아버지 최무선(崔茂宣)
[출처] 조선의 비밀 병기 : 총통기 화차(銃筒機火車)|작성자 도심

Jongwook Woo
CalStateLA
Data Issues
Cannot handle with the legacy approach
Too big
Non-/Semi-structured data
 3 Vs, 4 Vs,…
– Velocity, Volume, Variety
Traditional Systems can handle them
– But Again, Too expensive
Need new systems
Non-expensive

Jongwook Woo
CalStateLA
Two Cores in Big Data
How to store Big Data
How to compute Big Data
Google
How to store Big Data
– GFS
– Distributed Systems on non-expensive commodity computers
How to compute Big Data
– MapReduce
– Parallel Computing with non-expensive computers
Own super computers
Published papers in 2003, 2004

Jongwook Woo
CalStateLA
Super Computer vs Big Data vs Cloud
Traditional Super Computer
(Parallel File Systems: Lustre, PVFS, GPFS)
Cluster for Store
Big Data (Hadoop, Spark, Distributed Deep Learning)
Cluster for Compute and Store
(Distributed File Systems: HDFS, GFS)
However, Cloud Computing adopts
this separated architecture:
with High Speed N/W and Object
Storage
Cluster for Compute

Jongwook Woo
CalStateLA
Big Data: Hadoop
20
 Apache Hadoop Project in
Jan, 2006 split from Nutch
 Hadoop Founder:
o Doug Cutting
 Apache Committer:
Lucene, Nutch, …

Jongwook Woo
CalStateLA
Definition: Big Data [W13]
Non-expensive platform that is distributed parallel systems and
that can store a large scale data and process it in parallel
Hadoop
– Non-expensive Super Computer
– More public than the traditional super computers
• You can store and process your applications
– In your university labs, small companies, research centers
Others with storage and computing services
– Spark
• normally integrated into Hadoop with Hadoop community
– NoSQL DB (Cassandra, MongoDB, Redis, Hbase,…)
– ElasticSearch

Jongwook Woo
CalStateLA
Big Data: Linearly Scalable
 Some people questions that the system to handle 1 ~ 3GB of
data set is not Big Data
Well…. add more servers as more data in the future in Big Data platform
– it is linearly scalable once built
– n time more computing power ideally
Data Size: < 3 GB Data Size: 200 TB >
Add n
servers

Jongwook Woo
CalStateLA
Big Data Cluster
Are you ready for research now?
Large Scale Data Set with computing engine: ML, DS
Massive Data Set with
Computing Engines (Hadoop,
Spark)

Jongwook Woo
CalStateLA
Experimental Results in AWS [PMBW18]
Execution times
Big Data Science
3 nodes:
–40min – 70mins
11 nodes
–10min – 20mins

Jongwook Woo
CalStateLA
Big Data is great for Any Small Business
Your data is the value and Big Data
 Customer data
 Operational data
You have your specific data
Big Company does not have a specific data as you have
Potentials
 Your customer data
– Smart marketing and Sales
– Advertisement
 Your operational data
– Efficient operation, For Example, Smart*:
• Smart Factory, Smart City

Jongwook Woo
CalStateLA
Big Data Data Analysis & Visualization
Sentiment Map of Alphago
Positive
Negative

Jongwook Woo
CalStateLA
K-Election 2017
(April 29 – May 9)

Jongwook Woo
CalStateLA
IoT of Smart Factory
28

Jongwook Woo
CalStateLA
IoT of Smart Factory (Cont’d)
29

Jongwook Woo
CalStateLA
Businesses popular in 5 miles of CalStateLA,
USC , UCLA

Jongwook Woo
CalStateLA
Jams and other traffic incidents reported
by users in Dec 2017 – Jan 2018: [DW19a]

Jongwook Woo
CalStateLA
Contents
 Myself
 Introduction To Big Data
 Deep Learning and Big Data
 Big Data Predictive Analysis
Summary

Jongwook Woo
CalStateLA
Big Data Analysis and Prediction
Big Data Analysis
Hadoop, Spark, NoSQL DB, SAP HANA, ElasticSearch,..
Big Data for Data Analysis
– How to store, compute, analyze massive dataset?
Big Data Science
How to predict the future trend and pattern with the massive
dataset? => Machine Learning

Jongwook Woo
CalStateLA
Spark
 Limitation in MapReduce
 Hard to program in Java
 Batch Processing
– Not interactive
 Disk storage for intermediate data
– Performance issue
 Spark by UC Berkley AMP Lab
 Started by Matei Zaharia in 2009,
– and open sourced in 2010
In-Memory storage for intermediate data
 20 ~ 100 times faster than
– MapReduce
Good in Machine Learning => Big Data Science
– Iterative algorithms

Jongwook Woo
CalStateLA
Spark (Cont’d)
Spark ML
Supports Machine Learning libraries
Process massive data set to build prediction models

Jongwook Woo
CalStateLA
Big Data Analysis and Prediction Flow
Data Collection
Batch API: Yelp, Google
Streaming: Twitter, Apache
NiFi, Kafka, StereamSets,
Storm
Open Data: Government
Data Storage
HDFS, S3, Object Storage,
NoSQL DB (Couchbase)…
Data Filtering
Hive, Pig
Data Analysis and Science
Hive, Pig, Spark, Deep Learning,
BI Tools (Qlik, Tableau, …)
Data Visualization
Qlik, Excel PowerMap,
Tableau, Looker, …
- Engineering:
- Big Data Engineering
- Big Data Analysis
- Data Visualization
- Research
- Big Data Science Deep Learning

Jongwook Woo
CalStateLA
Traditional Data Science
The Gap
Big Data Engineers, Scientists, Analysts, etc.
Gap between Traditional Data Science and Big Data
Communities

Jongwook Woo
CalStateLA
Leveraging Big Data Cluster [MCSPBW19, DW19a]
 Existing Big Data cluster with massive data set with the
traditional ML
Issues and Solutions: Too
slow in large scale data
migration and single
server fails
Single server for
Python and R
Traditional Machine
Learning
Big Data Cluster

Jongwook Woo
CalStateLA
Deep Learning
 Machine Learning
 Has been popular since Google Tensorflow
 Multiple Cores in GPU
– Even with multiple GPUs and CPUs
 Parallel Computing
 GPU (Nvidia GTX 1660 Ti)
 1280 CUDA cores
 Deep Learning Libraries
 Tensor Flow
 PyTorch
 Keras
 Caffe, Caffe2
 Microsoft Cognitive Toolkit (Previously CNTK)
 Apache Mxnet
 DeepLearning4j
 …

Jongwook Woo
CalStateLA
From Neural Networks to Deep Learning
Deep learning – Different types of architectures
Generative Adversarial Networks (GAN)
Convolutional Neural Networks (CNN)
Neural Networks (NN)
7 © 2017 SAP SE or an SAP affiliate company. All rights
reserved. ǀ PUBLIC
Recurrent Neural Networks (RNN) &
Long-Short Term Memory (LSTM)
Ref: SAP Enterprise Deep Learning with TensorFlow

Jongwook Woo
CalStateLA
Deep Learning
CNN
Image Recognition
Video Analysis
 NLP for classification, Prediction
RNN
Time Series Prediction
Speech Recognition/Synthesis
Image/Video Captioning
Text Analysis
– Conversation Q&A
GAN
 Media Generation
– Photo Realistic Images
Human Image Synthesis: Fake faces

Jongwook Woo
CalStateLA
Data Scale Driving: Deep Learning Process
Deep Learning and Massive Data [3]
“Machine Learning Yearning” Andrew Ng 2016

Jongwook Woo
CalStateLA
Deep learning experts
The
Chasm
Big Data Engineers, Scientists, Analysts, etc.
Another Gap between Deep Learning and Big Data
Communities [6]

Jongwook Woo
CalStateLA
Leveraging Big Data Cluster
 Existing Big Data cluster with massive data set without using
Big Data
Too slow in data
migration and
single server fails
Single GPU
server for Deep
Learning?
Single server for
Python and R
Traditional
Machine Learning?
Big Data Cluster

Jongwook Woo
CalStateLA
Deep Learning with Spark
What if we combine Deep Learning and Spark?

Jongwook Woo
CalStateLA
Leveraging Big Data Cluster
 Existing Big Data cluster
Big Data Engineering
Big Data Analysis
Big Data Science
Distributed Deep Learning
– Integrate Deep Learning to the cluster
Not needs data migration and can leverage the
parallel computing and existing large scale data
Big Data Cluster

Jongwook Woo
CalStateLA
Deep Learning with Spark
Deep Learning Pipelines for Apache Spark
Databricks
TensorFlowOnSpark
Yahoo! Inc
BigDL (Distributed Deep Learning Library for Apache Spark)
Intel
DL4J (Deeplearning4j On Spark)
Skymind
Distributed Deep Learning with Keras & Spark
Elephas

Jongwook Woo
CalStateLA
Big Data Prediction with DDL
DDL: Distributed Deep Learning
Tensor Flow
Distributed Training and Inference in Spark cluster
DDL

Jongwook Woo
CalStateLA
Spark ML and DDL [MKW19]
Deep Learning in Spark cluster
DDL
DDL lib
DDL lib
Deep Learning in Spark

Jongwook Woo
CalStateLA
Azure ML Studio and Spark ML Result Comparison:
Ad Click Fraud Prediction, 7GB data [GLBW19]
TWO-CLASS
DECISION
JUNGLE
(AzureML)
TWO-CLASS
DECISION
FOREST
(AzureML)
DECISION
TREE
CLASSIFIER
(Databricks)
RANDOM
FOREST
CLASSIFIER
(Databricks)
DECISION TREE
CLASSIFIER
(Balanced
Sample Data,
Oracle)
RANDOM
FOREST
CLASSIFIER
(Balanced
Sample Data,
Oracle)
AUC 0.905 0.997 0.815 0.746 0.896 0.893
PRECISION 1.0 0.992 0.822 0.878 0.935 0.934
RECALL 0.001 0.902 0.633 0.495 0.807 0.800
TP 35 47,199 86,683 67,726 111,187 110,220
FP 0 377 18,727 9,408 7,712 7,791
TN 52,306 406,228 7,112,961 7,122,280 545,302 545,223
FN 406,605 5,142 50,074 69,031 26,604 27,571
Run Time 2 hrs 2-3 hrs 22 mins 50 mins 24 sec 2 mins

Jongwook Woo
CalStateLA
Big Data Science: Transaction Data Fraud Detection
[PMBW18]
Model Area under
ROC
Precision Recall
DecisionTreeClassifier
RandomForestClassifier 0.909573
LogisticRegression
 Size: 470 MB (=> 718MB)
 6,362,620 records
 Not that large scale data comparing to data set > GB
 https://www.kaggle.com/ntnu-testimon/paysim1
3 models in Spark Cluster with different combinations of the
parameters
 Times taken: 1 hour with 3 Spark clsters
 In theory of Linear Scalability: 2 minutes with 30 Spark clsters

Jongwook Woo
CalStateLA
Experimental Results in AWS [PMBW18]
Execution times
3 nodes:
–40min – 70mins
11 nodes
–10min – 20mins
Shows Scalability

Big Data AI Center (BigDAI / HiPIC)
Jongwook Woo
CalStateLA
Big Data Science in Smart * [DW19a]
 Traffic Data Analysis and Prediction using Big Data
Smart*
– Smart Things
• Data collected from Cellphone apps
– Traffic Data from the driver and the cell phone
Data source:
Navigation app traffic data set from LA City Department*
– Information reported by users – Alerts
information captured by user’s device – Jams
*Limited authorization to access the full datasets 100 GB + original;
– Adopted limited dataset to 9 days (Dec 31– Jan 8, 2018)
– ~2GB

Jongwook Woo
CalStateLA
Introduction
Provide real-time directions and up-to-date information
Traffic
Accidents
Road closure
Weather hazards
Lurking police vehicles and etc.
We are going to find out:
Areas with high volume of traffic (geography)
Peak-hours
Density of Alerts and Incidents
Traffic volume by road types
Prediction of traffic jam

Jongwook Woo
CalStateLA
Experiment Environment:
Traditional Systems and Big Data

Jongwook Woo
CalStateLA
H/W Specification
 Hadoop Spark Cluster
Number of nodes 6
OCPUs 12
CPU speed 2.2GHz
Memory 180 GB
Storage 682 GB

Jongwook Woo
CalStateLA
Implementation Flow
Big Data Science and AI ML
Local Computer
Raw data
files (JSON)
Geo-Spatial
Visualization (3D map)
Dashboard for Analytics
Big Data Analysis:
Hadoop/Hive
Upload dataset to
HDFS
Parse JSON files using
Pandas
Create tables’ schema
Clean data
Create sample/summary
dataset for prediction and
visualization
Traditional Data
Science: Microsoft
Azure ML Studio
Upload sample dataset
Apply data
transformation
Split dataset for
training and scoring
Train model(s)
Evaluate model(s)

Jongwook Woo
CalStateLA
Traffic Dashboard: Big Data Analysis
Peak
Peak

Jongwook Woo
CalStateLA
Traffic Dashboard: Big Data Analysis (Cont’d)
Major areas of traffic are:
Downtown Los Angeles
Santa Monica
Hollywood
Freeway (highways)

Jongwook Woo
CalStateLA
Video-Simulation of Traffic in LA (captured from users' devices)

Jongwook Woo
CalStateLA
Video-Simulation of Traffic in LA (reported by app users)

Jongwook Woo
CalStateLA
Features/columns in a dataset
location x, location y X and Y -coordinate of location
date_pst Pacific Time of the publication of traffic report
*date splits into month, day, hour, min, sec, weekday
speed driver’s captured speed in mph
length length of the traffic ahead in the route of user in meters
level jam level: 1 – 5
where (1: almost no jam) and (5: standstill jam)

Jongwook Woo
CalStateLA
MODEL Evaluation: Traditional Data Science
with Azure ML Studio
Model Accuracy Precision Recall AUC ROC
LR 0.662 0.662 1.0 0.571
BDT 0.805 0.832 0.884 0.868
DF 0.832 0.868 0.880 0.885

Jongwook Woo
CalStateLA
Summary of Traffic Prediction with
Machine Learning
Model is based on sampled
dataset ~ 1M rows (100 MB):
Sampled using Spark as the data set
is 2GB
Best model - Decision Forest
Accuracy – 0.832
Precision - 0.868
Recall - 0.880
Area under the Curve – 0.885
Confusion Matrix

Jongwook Woo
CalStateLA
Distributed Deep Learning in Big Data Cluster
[MKW19]
Predictive Analysis
Prediction of rating
– important measures for purchase and selling
Spark ML: ALS (Alternating Least Squares) algorithm
DDL (Distributed Deep Learning): Neural Collaborative Filtering (NCF)
Dataset : - https://s3.amazonaws.com/amazon-reviews-
pds/tsv/index.txt
Products reviewed between 2005 and 2015 are analyzed
Total product reviews : 9.57 million
File Size : 5.26 GB

Jongwook Woo
CalStateLA
Summary: Performance

Jongwook Woo
CalStateLA
Summary: Mean Absolute Error

Jongwook Woo
CalStateLA
Training and Education
Emerging Technology every moment
IT companies lead the industry not university
How to catch up with?
– Training and Education
Company with new technology
Always deliver training
– Big Data
• Cloudera, Hortonworks
– AI Deep Learning
• Traditional Concept
– Stanford, UC Berkeley, edx, IBM, H2O

Jongwook Woo
CalStateLA
Training (Cont’d)
Training by Company
 3 - 4days/Week
– $2,500 - $3,000
– Practical
• with theory + hands-on exercise
• Instructor paid well
• Employer send their engineers to learn the new technology in a few
weeks
Education in University
Need an instructor who knows the new technology
– Not easy
• IT companies lead the industry not university

Jongwook Woo
CalStateLA
Trained but No Experience with bad management in Korea
Sang-Ryung Battle:
From 2017 Korean
Blockbuster Movie,
“The Fortress”
(남한산성)

Jongwook Woo
CalStateLA
Trained Well With Experience and Good management in Japan
Battle of Nagashino,
1575, Japan

Jongwook Woo
CalStateLA
Trained but No Experience with bad management in Korea (Cont’d)
Sang-Ryung Battle:
From 2017 Korean
Blockbuster Movie,
“The Fortress”
(남한산성)

Jongwook Woo
CalStateLA
Summary
Introduction to Big Data
Definition in terms of platforms
Data and Predictive Analysis in Massive Data Set
Introduction to Deep Learning in Big Data
Big Data Predictive Analysis
Big Data Science
Education is important

Jongwook Woo
CalStateLA
Questions?

Jongwook Woo
CalStateLA
References
1. [W13] Jongwook Woo, DMKD-00150, “Market Basket Analysis Algorithms with MapReduce”, Wiley
Interdisciplinary Reviews Data Mining and Knowledge Discovery, Oct 28 2013, Volume 3, Issue 6, pp445-452, ISSN
1942-4795
2. [KCWW16] “Big Data Analysis using Spark for Collision Rate Near CalStateLA” , Manik Katyal, Parag Chhadva,
Shubhra Wahi & Jongwook Woo, https://globaljournals.org/GJCST_Volume16/1-Big-Data-Analysis-using-Spark.pdf
3. [PMBW18] Priyanka Purushu, Niklas Melcher, Bhagyashree Bhagwat, Jongwook Woo, "Predictive Analysis of
Financial Fraud Detection using Azure and Spark ML", Asia Pacific Journal of Information Systems (APJIS),
VOL.28│NO.4│December 2018, pp308~319
4. [MCSPBW19] Monika Mishra, Jaydeep Chopde, Maitri Shah, Pankti Parikh, Rakshith Chandan Babu, Jongwook Woo,
"Big Data Predictive Analysis of Amazon Product Review", KSII The 14th Asia Pacific International Conference on
Information Science and Technology (APIC-IST) 2019, pp141-147, ISSN 2093-0542
5. [GLBW19] Neha Gupta, Hai Anh Le, Maria Boldina, Jongwook Woo, "Predicting fraud of AD click using Traditional
and Spark ML", KSII The 14th Asia Pacific International Conference on Information Science and Technology (APIC-
IST) 2019, pp24-28, ISSN 2093-0542
6. [DW19a] Dalyapraz Dauletbak, Jongwook Woo, "Traffic Data Analysis and Prediction using Big Data", KSII The 14th
Asia Pacific International Conference on Information Science and Technology (APIC-IST) 2019, pp127-133, ISSN
2093-0542
7. [SW19] Ruchi Singh and Jongwook Woo, "Applications of Machine Learning Models on Yelp Data", Asia Pacific
Journal of Information Systems (APJIS), Vol.29, No.1, 2019, pp35-49, ISSN 2288-5404

Jongwook Woo
CalStateLA
References
8. [MKW19] Monika Mishra, Mingoo Kang, Jongwook Woo, “Rating Prediction using Deep Learning and Spark”, The 11th
International Conference on Internet (ICONI 2019), Dec 15-18 2019, Hanoi, Vietnam
9. [DW19b] (Will be Published Dec 2019) Dalyapraz Dauletbak, Jongwook Woo, “Big Data Analysis and Prediction of Traffic in
Los Angeles”, in Transactions on Internet & Information Systems (TIIS)
10. Which Is Deeper - Comparison Of Deep Learning Frameworks On Spark, https://www.slideshare.net/SparkSummit/which-
is-deeper-comparison-of-deep-learning-frameworks-on-spark
11. Accelerating Machine Learning and Deep Learning At Scale with Apache Spark,
https://www.slideshare.net/SparkSummit/accelerating-machine-learning-and-deep-learning-at-scalewith-apache-spark-
keynote-by-ziya-ma
12. Deep Learning with Apache Spark and TensorFlow, https://databricks.com/blog/2016/01/25/deep-learning-with-apache-
spark-and-tensorflow.html
13. Overview of Smart Factory, https://www.slideshare.net/BrendanSheppard1/overview-of-smart-factory-solutions-
68137094/6
14. TensorFrames: Google Tensorflow on Apache Spark, https://www.slideshare.net/databricks/tensorframes-google-
tensorflow-on-apache-spark
15. Deep learning and Apache Spark, https://www.slideshare.net/QuantUniversity/deep-learning-and-apache-spark

History and Trend of Big Data and Deep Learning

More Related Content

What's hot (20)

Similar to History and Trend of Big Data and Deep Learning (20)

More from Jongwook Woo (14)

Recently uploaded (20)

History and Trend of Big Data and Deep Learning